%PDF-1.3 1 0 obj << /Type /Catalog /Outlines 2 0 R /Pages 3 0 R >> endobj 2 0 obj << /Type /Outlines /Count 0 >> endobj 3 0 obj << /Type /Pages /Kids [6 0 R 236 0 R 335 0 R 355 0 R 357 0 R 373 0 R 395 0 R 439 0 R ] /Count 8 /Resources << /ProcSet 4 0 R /Font << /F1 8 0 R /F2 9 0 R /F3 10 0 R /F4 11 0 R /F5 238 0 R >> /XObject << /I1 12 0 R /I2 13 0 R /I3 257 0 R /I4 258 0 R /I5 261 0 R /I6 262 0 R /I7 263 0 R /I8 264 0 R /I9 265 0 R /I10 266 0 R /I11 267 0 R /I12 268 0 R /I13 269 0 R /I14 270 0 R /I15 289 0 R /I16 290 0 R /I17 293 0 R /I18 294 0 R /I19 295 0 R /I20 296 0 R /I21 297 0 R /I22 298 0 R /I23 299 0 R /I24 300 0 R /I25 301 0 R /I26 302 0 R /I27 321 0 R /I28 322 0 R /I29 325 0 R /I30 326 0 R /I31 327 0 R /I32 328 0 R /I33 329 0 R /I34 330 0 R /I35 331 0 R /I36 332 0 R /I37 333 0 R /I38 334 0 R /I39 337 0 R /I40 338 0 R /I41 343 0 R /I42 344 0 R /I43 349 0 R /I44 350 0 R /I45 361 0 R /I46 362 0 R /I47 365 0 R /I48 366 0 R /I49 369 0 R /I50 370 0 R /I51 371 0 R /I52 372 0 R /I53 377 0 R /I54 378 0 R /I55 383 0 R /I56 384 0 R /I57 389 0 R /I58 390 0 R /I59 393 0 R /I60 394 0 R >> >> /MediaBox [0.000 0.000 612.000 792.000] >> endobj 4 0 obj [/PDF /Text /ImageC ] endobj 5 0 obj << /Creator (DOMPDF) /CreationDate (D:20180722224038+00'00') /ModDate (D:20180722224038+00'00') /Title (Twitter Improves Influenza Forecasting PLOS Currents Outbreaks) >> endobj 6 0 obj << /Type /Page /Parent 3 0 R /Annots [ 14 0 R 16 0 R 18 0 R 20 0 R 22 0 R 24 0 R 26 0 R 28 0 R 30 0 R 32 0 R 34 0 R 36 0 R 38 0 R 40 0 R 42 0 R 44 0 R 46 0 R 48 0 R 50 0 R 52 0 R 54 0 R 56 0 R 58 0 R 60 0 R 62 0 R 64 0 R 66 0 R 68 0 R 70 0 R 72 0 R 74 0 R 76 0 R 78 0 R 80 0 R 82 0 R 84 0 R 86 0 R 88 0 R 90 0 R 92 0 R 94 0 R 96 0 R 98 0 R 100 0 R 102 0 R 104 0 R 106 0 R 108 0 R 110 0 R 112 0 R 114 0 R 116 0 R 118 0 R 120 0 R 122 0 R 124 0 R 126 0 R 128 0 R 130 0 R 132 0 R 134 0 R 136 0 R 138 0 R 140 0 R 142 0 R 144 0 R 146 0 R 148 0 R 150 0 R 152 0 R 154 0 R 156 0 R 158 0 R 160 0 R 162 0 R 164 0 R 166 0 R 168 0 R 170 0 R 172 0 R 174 0 R 176 0 R 178 0 R 180 0 R 182 0 R 184 0 R 186 0 R 188 0 R 190 0 R 192 0 R 194 0 R 196 0 R 198 0 R 200 0 R 202 0 R 204 0 R 206 0 R 208 0 R 210 0 R 212 0 R 214 0 R 216 0 R 218 0 R 220 0 R 222 0 R 224 0 R 226 0 R 228 0 R 230 0 R 232 0 R 234 0 R ] /Contents 7 0 R >> endobj 7 0 obj << /Length 34838 >> stream q 375.000 0 0 39.000 222.000 738.000 cm /I2 Do Q q 15.000 709.302 577.500 28.698 re W n 0.267 0.267 0.267 rg BT 15.000 718.042 Td /F2 21.0 Tf [(Twitter Improves Influenza Forecasting)] TJ ET Q 0.271 0.267 0.267 rg BT 15.000 700.036 Td /F3 9.8 Tf [(October 28, 2014)] TJ ET BT 88.388 700.036 Td /F3 9.8 Tf [()] TJ ET 0.267 0.267 0.267 rg BT 93.263 700.036 Td /F3 9.8 Tf [(Research Article)] TJ ET BT 26.250 688.195 Td /F1 9.8 Tf [(Michael J. Paul)] TJ ET 0.271 0.267 0.267 rg BT 92.355 688.195 Td /F1 9.8 Tf [(, )] TJ ET 0.267 0.267 0.267 rg BT 97.776 688.195 Td /F1 9.8 Tf [(Mark Dredze)] TJ ET 0.271 0.267 0.267 rg BT 153.575 688.195 Td /F1 9.8 Tf [(, )] TJ ET 0.267 0.267 0.267 rg BT 158.996 688.195 Td /F1 9.8 Tf [(David Broniatowski)] TJ ET 0.271 0.267 0.267 rg BT 26.250 676.290 Td /F1 9.8 Tf [(Paul MJ, Dredze M, Broniatowski D. Twitter Improves Influenza Forecasting. PLOS Currents Outbreaks. 2014 Oct 28 . Edition )] TJ ET BT 26.250 664.386 Td /F1 9.8 Tf [(1. doi: 10.1371/currents.outbreaks.90b9ed0f59bae4ccaa683a39865d9117.)] TJ ET q 15.000 41.264 577.500 620.741 re W n 0.271 0.267 0.267 rg BT 26.250 635.283 Td /F4 12.0 Tf [(Abstract)] TJ ET BT 26.250 615.329 Td /F1 9.8 Tf [(Accurate disease forecasts are imperative when preparing for influenza epidemic outbreaks; nevertheless, these forecasts are )] TJ ET BT 26.250 603.424 Td /F1 9.8 Tf [(often limited by the time required to collect new, accurate data. In this paper, we show that data from the microblogging )] TJ ET BT 26.250 591.519 Td /F1 9.8 Tf [(community Twitter significantly improves influenza forecasting. Most prior influenza forecast models are tested against historical )] TJ ET BT 26.250 579.615 Td /F1 9.8 Tf [(influenza-like illness \(ILI\) data from the U.S. Centers for Disease Control and Prevention \(CDC\). These data are released with a )] TJ ET BT 26.250 567.710 Td /F1 9.8 Tf [(one-week lag and are often initially inaccurate until the CDC revises them weeks later. Since previous studies utilize the final, )] TJ ET BT 26.250 555.805 Td /F1 9.8 Tf [(revised data in evaluation, their evaluations do not properly determine the effectiveness of forecasting. Our experiments using )] TJ ET BT 26.250 543.900 Td /F1 9.8 Tf [(ILI data available at the time of the forecast show that models incorporating data derived from Twitter can reduce forecasting )] TJ ET BT 26.250 531.996 Td /F1 9.8 Tf [(error by 17-30% over a baseline that only uses historical data. For a given level of accuracy, using Twitter data produces )] TJ ET BT 26.250 520.091 Td /F1 9.8 Tf [(forecasts that are two to four weeks ahead of baseline models. Additionally, we find that models using Twitter data are, on )] TJ ET BT 26.250 508.186 Td /F1 9.8 Tf [(average, better predictors of influenza prevalence than are models using data from Google Flu Trends, the leading web data )] TJ ET BT 26.250 496.281 Td /F1 9.8 Tf [(source.)] TJ ET BT 26.250 459.679 Td /F4 12.0 Tf [(Funding Statement)] TJ ET BT 26.250 439.725 Td /F1 9.8 Tf [(Mr. Paul was supported by a PhD fellowship from Microsoft Research. The funders had no role in study design, data collection )] TJ ET BT 26.250 427.820 Td /F1 9.8 Tf [(and analysis, decision to publish, or preparation of the manuscript.)] TJ ET BT 26.250 398.717 Td /F4 12.0 Tf [(Introduction)] TJ ET BT 26.250 378.763 Td /F1 9.8 Tf [(Accurate disease forecasts are imperative when preparing for influenza epidemic outbreaks. )] TJ ET 0.267 0.267 0.267 rg BT 426.166 380.270 Td /F4 8.7 Tf [(1)] TJ ET 0.271 0.267 0.267 rg BT 430.984 382.651 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 433.394 380.270 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 438.212 378.763 Td /F1 9.8 Tf [( This need has driven the )] TJ ET BT 26.250 366.858 Td /F1 9.8 Tf [(research community to bring a multitude of influenza forecasting methods to bear, drawing upon a wide range of statistical )] TJ ET BT 26.250 354.954 Td /F1 9.8 Tf [(techniques and laboratory, clinical, epidemiological, climatological, and demographic data sources. )] TJ ET 0.267 0.267 0.267 rg BT 454.343 356.461 Td /F4 8.7 Tf [(1)] TJ ET 0.271 0.267 0.267 rg BT 459.162 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 461.571 356.461 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 466.390 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 468.799 356.461 Td /F4 8.7 Tf [(3)] TJ ET 0.271 0.267 0.267 rg BT 473.618 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 476.027 356.461 Td /F4 8.7 Tf [(4)] TJ ET 0.271 0.267 0.267 rg BT 480.846 354.954 Td /F1 9.8 Tf [( Nevertheless, )] TJ ET BT 26.250 343.049 Td /F1 9.8 Tf [(disease forecasts are often limited by the time required to collect new, accurate data.)] TJ ET BT 26.250 323.644 Td /F1 9.8 Tf [(Recent work has drawn upon novel web data especially Twitter )] TJ ET 0.267 0.267 0.267 rg BT 309.117 325.151 Td /F4 8.7 Tf [(5)] TJ ET 0.271 0.267 0.267 rg BT 313.936 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 316.345 325.151 Td /F4 8.7 Tf [(6)] TJ ET 0.271 0.267 0.267 rg BT 321.164 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 323.573 325.151 Td /F4 8.7 Tf [(7)] TJ ET 0.271 0.267 0.267 rg BT 328.392 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 330.801 325.151 Td /F4 8.7 Tf [(8)] TJ ET 0.271 0.267 0.267 rg BT 335.620 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 338.029 325.151 Td /F4 8.7 Tf [(9)] TJ ET 0.271 0.267 0.267 rg BT 342.848 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 345.257 325.151 Td /F4 8.7 Tf [(10)] TJ ET 0.271 0.267 0.267 rg BT 354.894 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 357.304 325.151 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 366.941 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 369.350 325.151 Td /F4 8.7 Tf [(12)] TJ ET 0.271 0.267 0.267 rg BT 378.988 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 381.397 325.151 Td /F4 8.7 Tf [(29)] TJ ET 0.271 0.267 0.267 rg BT 391.034 323.644 Td /F1 9.8 Tf [( messages and Google search queries )] TJ ET 0.267 0.267 0.267 rg BT 560.655 325.151 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 570.292 323.644 Td /F1 9.8 Tf [( )] TJ ET BT 26.250 311.739 Td /F1 9.8 Tf [(in order to detect influenza rates in real time \(i.e., influenza surveillance )] TJ ET 0.267 0.267 0.267 rg BT 336.212 313.247 Td /F4 8.7 Tf [(14)] TJ ET 0.271 0.267 0.267 rg BT 345.850 315.628 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 348.259 313.247 Td /F4 8.7 Tf [(15)] TJ ET 0.271 0.267 0.267 rg BT 357.896 315.628 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 360.306 313.247 Td /F4 8.7 Tf [(16)] TJ ET 0.271 0.267 0.267 rg BT 369.943 311.739 Td /F1 9.8 Tf [(\). Although Google Flu Trends \(GFT\) has )] TJ ET BT 26.250 299.835 Td /F1 9.8 Tf [(demonstrated some forecast accuracy, )] TJ ET 0.267 0.267 0.267 rg BT 196.943 301.342 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 201.762 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 204.171 301.342 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 213.809 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 216.218 301.342 Td /F4 8.7 Tf [(18)] TJ ET 0.271 0.267 0.267 rg BT 225.855 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 228.265 301.342 Td /F4 8.7 Tf [(19)] TJ ET 0.271 0.267 0.267 rg BT 237.902 299.835 Td /F1 9.8 Tf [( it has recently been criticized because of its sensitivity to media reports, the )] TJ ET BT 26.250 287.930 Td /F1 9.8 Tf [(lack of transparency behind GFT data, and the infrequency with which GFT models are updated. )] TJ ET 0.267 0.267 0.267 rg BT 444.613 289.437 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 454.250 291.818 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 456.659 289.437 Td /F4 8.7 Tf [(20)] TJ ET 0.271 0.267 0.267 rg BT 466.297 287.930 Td /F1 9.8 Tf [( In contrast, the )] TJ ET BT 26.250 276.025 Td /F1 9.8 Tf [(forecasting potential of open social media, and Twitter in particular, remains largely untested.)] TJ ET BT 26.250 256.620 Td /F1 9.8 Tf [(In this paper, we demonstrate that influenza surveillance signals from Twitter significantly improve influenza forecasting. We use )] TJ ET BT 26.250 244.716 Td /F1 9.8 Tf [(freely available Twitter data and methods that are insensitive to influence from the mainstream media. )] TJ ET 0.267 0.267 0.267 rg BT 467.350 246.223 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 476.987 248.604 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 479.396 246.223 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 489.034 244.716 Td /F1 9.8 Tf [( We are the first to )] TJ ET BT 26.250 232.811 Td /F1 9.8 Tf [(perform an explicit forecast of influenza prevalence rates weeks into the future using social media data, and the first to compare )] TJ ET BT 26.250 220.906 Td /F1 9.8 Tf [(social media to GFT.)] TJ ET 0.267 0.267 0.267 rg BT 116.194 222.413 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 125.831 224.794 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 128.240 222.413 Td /F4 8.7 Tf [(30)] TJ ET 0.271 0.267 0.267 rg BT 26.250 201.501 Td /F1 9.8 Tf [(Our findings indicate that Twitter data are both more accessible, and provide better forecasts, when compared to GFT data. )] TJ ET 0.267 0.267 0.267 rg BT 560.579 203.009 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 26.250 189.597 Td /F1 9.8 Tf [(This is an important validation of social media data sources for influenza surveillance and forecasting.)] TJ ET BT 26.250 170.192 Td /F1 9.8 Tf [(Our analysis is also the first to use historically accurate surveillance data for the United States. Prior work has relied upon )] TJ ET BT 26.250 158.287 Td /F1 9.8 Tf [(amended data that were not available at the time the forecast was required. These data, consisting of Outpatient Influenza-like )] TJ ET BT 26.250 146.382 Td /F1 9.8 Tf [(Illness Surveillance Network \(ILINet\) reports from the U.S. Centers for Disease Control and Prevention, are the gold standard )] TJ ET BT 26.250 134.478 Td /F1 9.8 Tf [(for United States influenza surveillance. ILINet data are published with a one-week lag, though some have interpreted the lag as )] TJ ET BT 26.250 122.573 Td /F1 9.8 Tf [(being two weeks, )] TJ ET 0.267 0.267 0.267 rg BT 103.743 124.080 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 113.380 122.573 Td /F1 9.8 Tf [( the difference depending the dates used in determining the lag \(first day of the reported week vs. the last.\) )] TJ ET BT 26.250 110.668 Td /F1 9.8 Tf [(Importantly, the numbers initially released in CDC reports are subject to future revisions as data from additional ILINet sentinel )] TJ ET BT 26.250 98.763 Td /F1 9.8 Tf [(sites arrive. Retrospective analyses of ILINet data generally rely upon the final numbers released by the CDC; not the data )] TJ ET BT 26.250 86.859 Td /F1 9.8 Tf [(initially available. The degree to which updates to ILINet data might impact forecast accuracy has not been previously )] TJ ET BT 26.250 74.954 Td /F1 9.8 Tf [(considered. \(The effect of revisions for forecasting in Latin America was recently examined in )] TJ ET 0.267 0.267 0.267 rg BT 429.949 76.461 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 439.586 74.954 Td /F1 9.8 Tf [( .\) Our results demonstrate that )] TJ ET BT 26.250 63.049 Td /F1 9.8 Tf [(these revisions make a significant difference in forecasting efficacy, further highlighting the benefits of using real-time social )] TJ ET BT 26.250 51.144 Td /F1 9.8 Tf [(media data such as Twitter.)] TJ ET Q q 15.000 709.302 577.500 28.698 re W n 0.267 0.267 0.267 rg BT 15.000 718.042 Td /F2 21.0 Tf [(Twitter Improves Influenza Forecasting)] TJ ET Q 0.271 0.267 0.267 rg BT 15.000 700.036 Td /F3 9.8 Tf [(October 28, 2014)] TJ ET BT 88.388 700.036 Td /F3 9.8 Tf [()] TJ ET 0.267 0.267 0.267 rg BT 93.263 700.036 Td /F3 9.8 Tf [(Research Article)] TJ ET BT 26.250 688.195 Td /F1 9.8 Tf [(Michael J. Paul)] TJ ET 0.271 0.267 0.267 rg BT 92.355 688.195 Td /F1 9.8 Tf [(, )] TJ ET 0.267 0.267 0.267 rg BT 97.776 688.195 Td /F1 9.8 Tf [(Mark Dredze)] TJ ET 0.271 0.267 0.267 rg BT 153.575 688.195 Td /F1 9.8 Tf [(, )] TJ ET 0.267 0.267 0.267 rg BT 158.996 688.195 Td /F1 9.8 Tf [(David Broniatowski)] TJ ET 0.271 0.267 0.267 rg BT 26.250 676.290 Td /F1 9.8 Tf [(Paul MJ, Dredze M, Broniatowski D. Twitter Improves Influenza Forecasting. PLOS Currents Outbreaks. 2014 Oct 28 . Edition )] TJ ET BT 26.250 664.386 Td /F1 9.8 Tf [(1. doi: 10.1371/currents.outbreaks.90b9ed0f59bae4ccaa683a39865d9117.)] TJ ET q 15.000 41.264 577.500 620.741 re W n 0.271 0.267 0.267 rg BT 26.250 635.283 Td /F4 12.0 Tf [(Abstract)] TJ ET BT 26.250 615.329 Td /F1 9.8 Tf [(Accurate disease forecasts are imperative when preparing for influenza epidemic outbreaks; nevertheless, these forecasts are )] TJ ET BT 26.250 603.424 Td /F1 9.8 Tf [(often limited by the time required to collect new, accurate data. In this paper, we show that data from the microblogging )] TJ ET BT 26.250 591.519 Td /F1 9.8 Tf [(community Twitter significantly improves influenza forecasting. Most prior influenza forecast models are tested against historical )] TJ ET BT 26.250 579.615 Td /F1 9.8 Tf [(influenza-like illness \(ILI\) data from the U.S. Centers for Disease Control and Prevention \(CDC\). These data are released with a )] TJ ET BT 26.250 567.710 Td /F1 9.8 Tf [(one-week lag and are often initially inaccurate until the CDC revises them weeks later. Since previous studies utilize the final, )] TJ ET BT 26.250 555.805 Td /F1 9.8 Tf [(revised data in evaluation, their evaluations do not properly determine the effectiveness of forecasting. Our experiments using )] TJ ET BT 26.250 543.900 Td /F1 9.8 Tf [(ILI data available at the time of the forecast show that models incorporating data derived from Twitter can reduce forecasting )] TJ ET BT 26.250 531.996 Td /F1 9.8 Tf [(error by 17-30% over a baseline that only uses historical data. For a given level of accuracy, using Twitter data produces )] TJ ET BT 26.250 520.091 Td /F1 9.8 Tf [(forecasts that are two to four weeks ahead of baseline models. Additionally, we find that models using Twitter data are, on )] TJ ET BT 26.250 508.186 Td /F1 9.8 Tf [(average, better predictors of influenza prevalence than are models using data from Google Flu Trends, the leading web data )] TJ ET BT 26.250 496.281 Td /F1 9.8 Tf [(source.)] TJ ET BT 26.250 459.679 Td /F4 12.0 Tf [(Funding Statement)] TJ ET BT 26.250 439.725 Td /F1 9.8 Tf [(Mr. Paul was supported by a PhD fellowship from Microsoft Research. The funders had no role in study design, data collection )] TJ ET BT 26.250 427.820 Td /F1 9.8 Tf [(and analysis, decision to publish, or preparation of the manuscript.)] TJ ET BT 26.250 398.717 Td /F4 12.0 Tf [(Introduction)] TJ ET BT 26.250 378.763 Td /F1 9.8 Tf [(Accurate disease forecasts are imperative when preparing for influenza epidemic outbreaks. )] TJ ET 0.267 0.267 0.267 rg BT 426.166 380.270 Td /F4 8.7 Tf [(1)] TJ ET 0.271 0.267 0.267 rg BT 430.984 382.651 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 433.394 380.270 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 438.212 378.763 Td /F1 9.8 Tf [( This need has driven the )] TJ ET BT 26.250 366.858 Td /F1 9.8 Tf [(research community to bring a multitude of influenza forecasting methods to bear, drawing upon a wide range of statistical )] TJ ET BT 26.250 354.954 Td /F1 9.8 Tf [(techniques and laboratory, clinical, epidemiological, climatological, and demographic data sources. )] TJ ET 0.267 0.267 0.267 rg BT 454.343 356.461 Td /F4 8.7 Tf [(1)] TJ ET 0.271 0.267 0.267 rg BT 459.162 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 461.571 356.461 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 466.390 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 468.799 356.461 Td /F4 8.7 Tf [(3)] TJ ET 0.271 0.267 0.267 rg BT 473.618 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 476.027 356.461 Td /F4 8.7 Tf [(4)] TJ ET 0.271 0.267 0.267 rg BT 480.846 354.954 Td /F1 9.8 Tf [( Nevertheless, )] TJ ET BT 26.250 343.049 Td /F1 9.8 Tf [(disease forecasts are often limited by the time required to collect new, accurate data.)] TJ ET BT 26.250 323.644 Td /F1 9.8 Tf [(Recent work has drawn upon novel web data especially Twitter )] TJ ET 0.267 0.267 0.267 rg BT 309.117 325.151 Td /F4 8.7 Tf [(5)] TJ ET 0.271 0.267 0.267 rg BT 313.936 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 316.345 325.151 Td /F4 8.7 Tf [(6)] TJ ET 0.271 0.267 0.267 rg BT 321.164 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 323.573 325.151 Td /F4 8.7 Tf [(7)] TJ ET 0.271 0.267 0.267 rg BT 328.392 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 330.801 325.151 Td /F4 8.7 Tf [(8)] TJ ET 0.271 0.267 0.267 rg BT 335.620 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 338.029 325.151 Td /F4 8.7 Tf [(9)] TJ ET 0.271 0.267 0.267 rg BT 342.848 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 345.257 325.151 Td /F4 8.7 Tf [(10)] TJ ET 0.271 0.267 0.267 rg BT 354.894 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 357.304 325.151 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 366.941 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 369.350 325.151 Td /F4 8.7 Tf [(12)] TJ ET 0.271 0.267 0.267 rg BT 378.988 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 381.397 325.151 Td /F4 8.7 Tf [(29)] TJ ET 0.271 0.267 0.267 rg BT 391.034 323.644 Td /F1 9.8 Tf [( messages and Google search queries )] TJ ET 0.267 0.267 0.267 rg BT 560.655 325.151 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 570.292 323.644 Td /F1 9.8 Tf [( )] TJ ET BT 26.250 311.739 Td /F1 9.8 Tf [(in order to detect influenza rates in real time \(i.e., influenza surveillance )] TJ ET 0.267 0.267 0.267 rg BT 336.212 313.247 Td /F4 8.7 Tf [(14)] TJ ET 0.271 0.267 0.267 rg BT 345.850 315.628 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 348.259 313.247 Td /F4 8.7 Tf [(15)] TJ ET 0.271 0.267 0.267 rg BT 357.896 315.628 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 360.306 313.247 Td /F4 8.7 Tf [(16)] TJ ET 0.271 0.267 0.267 rg BT 369.943 311.739 Td /F1 9.8 Tf [(\). Although Google Flu Trends \(GFT\) has )] TJ ET BT 26.250 299.835 Td /F1 9.8 Tf [(demonstrated some forecast accuracy, )] TJ ET 0.267 0.267 0.267 rg BT 196.943 301.342 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 201.762 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 204.171 301.342 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 213.809 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 216.218 301.342 Td /F4 8.7 Tf [(18)] TJ ET 0.271 0.267 0.267 rg BT 225.855 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 228.265 301.342 Td /F4 8.7 Tf [(19)] TJ ET 0.271 0.267 0.267 rg BT 237.902 299.835 Td /F1 9.8 Tf [( it has recently been criticized because of its sensitivity to media reports, the )] TJ ET BT 26.250 287.930 Td /F1 9.8 Tf [(lack of transparency behind GFT data, and the infrequency with which GFT models are updated. )] TJ ET 0.267 0.267 0.267 rg BT 444.613 289.437 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 454.250 291.818 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 456.659 289.437 Td /F4 8.7 Tf [(20)] TJ ET 0.271 0.267 0.267 rg BT 466.297 287.930 Td /F1 9.8 Tf [( In contrast, the )] TJ ET BT 26.250 276.025 Td /F1 9.8 Tf [(forecasting potential of open social media, and Twitter in particular, remains largely untested.)] TJ ET BT 26.250 256.620 Td /F1 9.8 Tf [(In this paper, we demonstrate that influenza surveillance signals from Twitter significantly improve influenza forecasting. We use )] TJ ET BT 26.250 244.716 Td /F1 9.8 Tf [(freely available Twitter data and methods that are insensitive to influence from the mainstream media. )] TJ ET 0.267 0.267 0.267 rg BT 467.350 246.223 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 476.987 248.604 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 479.396 246.223 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 489.034 244.716 Td /F1 9.8 Tf [( We are the first to )] TJ ET BT 26.250 232.811 Td /F1 9.8 Tf [(perform an explicit forecast of influenza prevalence rates weeks into the future using social media data, and the first to compare )] TJ ET BT 26.250 220.906 Td /F1 9.8 Tf [(social media to GFT.)] TJ ET 0.267 0.267 0.267 rg BT 116.194 222.413 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 125.831 224.794 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 128.240 222.413 Td /F4 8.7 Tf [(30)] TJ ET 0.271 0.267 0.267 rg BT 26.250 201.501 Td /F1 9.8 Tf [(Our findings indicate that Twitter data are both more accessible, and provide better forecasts, when compared to GFT data. )] TJ ET 0.267 0.267 0.267 rg BT 560.579 203.009 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 26.250 189.597 Td /F1 9.8 Tf [(This is an important validation of social media data sources for influenza surveillance and forecasting.)] TJ ET BT 26.250 170.192 Td /F1 9.8 Tf [(Our analysis is also the first to use historically accurate surveillance data for the United States. Prior work has relied upon )] TJ ET BT 26.250 158.287 Td /F1 9.8 Tf [(amended data that were not available at the time the forecast was required. These data, consisting of Outpatient Influenza-like )] TJ ET BT 26.250 146.382 Td /F1 9.8 Tf [(Illness Surveillance Network \(ILINet\) reports from the U.S. Centers for Disease Control and Prevention, are the gold standard )] TJ ET BT 26.250 134.478 Td /F1 9.8 Tf [(for United States influenza surveillance. ILINet data are published with a one-week lag, though some have interpreted the lag as )] TJ ET BT 26.250 122.573 Td /F1 9.8 Tf [(being two weeks, )] TJ ET 0.267 0.267 0.267 rg BT 103.743 124.080 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 113.380 122.573 Td /F1 9.8 Tf [( the difference depending the dates used in determining the lag \(first day of the reported week vs. the last.\) )] TJ ET BT 26.250 110.668 Td /F1 9.8 Tf [(Importantly, the numbers initially released in CDC reports are subject to future revisions as data from additional ILINet sentinel )] TJ ET BT 26.250 98.763 Td /F1 9.8 Tf [(sites arrive. Retrospective analyses of ILINet data generally rely upon the final numbers released by the CDC; not the data )] TJ ET BT 26.250 86.859 Td /F1 9.8 Tf [(initially available. The degree to which updates to ILINet data might impact forecast accuracy has not been previously )] TJ ET BT 26.250 74.954 Td /F1 9.8 Tf [(considered. \(The effect of revisions for forecasting in Latin America was recently examined in )] TJ ET 0.267 0.267 0.267 rg BT 429.949 76.461 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 439.586 74.954 Td /F1 9.8 Tf [( .\) Our results demonstrate that )] TJ ET BT 26.250 63.049 Td /F1 9.8 Tf [(these revisions make a significant difference in forecasting efficacy, further highlighting the benefits of using real-time social )] TJ ET BT 26.250 51.144 Td /F1 9.8 Tf [(media data such as Twitter.)] TJ ET Q q 15.000 709.302 577.500 28.698 re W n 0.267 0.267 0.267 rg BT 15.000 718.042 Td /F2 21.0 Tf [(Twitter Improves Influenza Forecasting)] TJ ET Q 0.271 0.267 0.267 rg BT 15.000 700.036 Td /F3 9.8 Tf [(October 28, 2014)] TJ ET BT 88.388 700.036 Td /F3 9.8 Tf [()] TJ ET 0.267 0.267 0.267 rg BT 93.263 700.036 Td /F3 9.8 Tf [(Research Article)] TJ ET BT 26.250 688.195 Td /F1 9.8 Tf [(Michael J. Paul)] TJ ET 0.271 0.267 0.267 rg BT 92.355 688.195 Td /F1 9.8 Tf [(, )] TJ ET 0.267 0.267 0.267 rg BT 97.776 688.195 Td /F1 9.8 Tf [(Mark Dredze)] TJ ET 0.271 0.267 0.267 rg BT 153.575 688.195 Td /F1 9.8 Tf [(, )] TJ ET 0.267 0.267 0.267 rg BT 158.996 688.195 Td /F1 9.8 Tf [(David Broniatowski)] TJ ET 0.271 0.267 0.267 rg BT 26.250 676.290 Td /F1 9.8 Tf [(Paul MJ, Dredze M, Broniatowski D. Twitter Improves Influenza Forecasting. PLOS Currents Outbreaks. 2014 Oct 28 . Edition )] TJ ET BT 26.250 664.386 Td /F1 9.8 Tf [(1. doi: 10.1371/currents.outbreaks.90b9ed0f59bae4ccaa683a39865d9117.)] TJ ET q 15.000 41.264 577.500 620.741 re W n 0.271 0.267 0.267 rg BT 26.250 635.283 Td /F4 12.0 Tf [(Abstract)] TJ ET BT 26.250 615.329 Td /F1 9.8 Tf [(Accurate disease forecasts are imperative when preparing for influenza epidemic outbreaks; nevertheless, these forecasts are )] TJ ET BT 26.250 603.424 Td /F1 9.8 Tf [(often limited by the time required to collect new, accurate data. In this paper, we show that data from the microblogging )] TJ ET BT 26.250 591.519 Td /F1 9.8 Tf [(community Twitter significantly improves influenza forecasting. Most prior influenza forecast models are tested against historical )] TJ ET BT 26.250 579.615 Td /F1 9.8 Tf [(influenza-like illness \(ILI\) data from the U.S. Centers for Disease Control and Prevention \(CDC\). These data are released with a )] TJ ET BT 26.250 567.710 Td /F1 9.8 Tf [(one-week lag and are often initially inaccurate until the CDC revises them weeks later. Since previous studies utilize the final, )] TJ ET BT 26.250 555.805 Td /F1 9.8 Tf [(revised data in evaluation, their evaluations do not properly determine the effectiveness of forecasting. Our experiments using )] TJ ET BT 26.250 543.900 Td /F1 9.8 Tf [(ILI data available at the time of the forecast show that models incorporating data derived from Twitter can reduce forecasting )] TJ ET BT 26.250 531.996 Td /F1 9.8 Tf [(error by 17-30% over a baseline that only uses historical data. For a given level of accuracy, using Twitter data produces )] TJ ET BT 26.250 520.091 Td /F1 9.8 Tf [(forecasts that are two to four weeks ahead of baseline models. Additionally, we find that models using Twitter data are, on )] TJ ET BT 26.250 508.186 Td /F1 9.8 Tf [(average, better predictors of influenza prevalence than are models using data from Google Flu Trends, the leading web data )] TJ ET BT 26.250 496.281 Td /F1 9.8 Tf [(source.)] TJ ET BT 26.250 459.679 Td /F4 12.0 Tf [(Funding Statement)] TJ ET BT 26.250 439.725 Td /F1 9.8 Tf [(Mr. Paul was supported by a PhD fellowship from Microsoft Research. The funders had no role in study design, data collection )] TJ ET BT 26.250 427.820 Td /F1 9.8 Tf [(and analysis, decision to publish, or preparation of the manuscript.)] TJ ET BT 26.250 398.717 Td /F4 12.0 Tf [(Introduction)] TJ ET BT 26.250 378.763 Td /F1 9.8 Tf [(Accurate disease forecasts are imperative when preparing for influenza epidemic outbreaks. )] TJ ET 0.267 0.267 0.267 rg BT 426.166 380.270 Td /F4 8.7 Tf [(1)] TJ ET 0.271 0.267 0.267 rg BT 430.984 382.651 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 433.394 380.270 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 438.212 378.763 Td /F1 9.8 Tf [( This need has driven the )] TJ ET BT 26.250 366.858 Td /F1 9.8 Tf [(research community to bring a multitude of influenza forecasting methods to bear, drawing upon a wide range of statistical )] TJ ET BT 26.250 354.954 Td /F1 9.8 Tf [(techniques and laboratory, clinical, epidemiological, climatological, and demographic data sources. )] TJ ET 0.267 0.267 0.267 rg BT 454.343 356.461 Td /F4 8.7 Tf [(1)] TJ ET 0.271 0.267 0.267 rg BT 459.162 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 461.571 356.461 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 466.390 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 468.799 356.461 Td /F4 8.7 Tf [(3)] TJ ET 0.271 0.267 0.267 rg BT 473.618 358.842 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 476.027 356.461 Td /F4 8.7 Tf [(4)] TJ ET 0.271 0.267 0.267 rg BT 480.846 354.954 Td /F1 9.8 Tf [( Nevertheless, )] TJ ET BT 26.250 343.049 Td /F1 9.8 Tf [(disease forecasts are often limited by the time required to collect new, accurate data.)] TJ ET BT 26.250 323.644 Td /F1 9.8 Tf [(Recent work has drawn upon novel web data especially Twitter )] TJ ET 0.267 0.267 0.267 rg BT 309.117 325.151 Td /F4 8.7 Tf [(5)] TJ ET 0.271 0.267 0.267 rg BT 313.936 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 316.345 325.151 Td /F4 8.7 Tf [(6)] TJ ET 0.271 0.267 0.267 rg BT 321.164 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 323.573 325.151 Td /F4 8.7 Tf [(7)] TJ ET 0.271 0.267 0.267 rg BT 328.392 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 330.801 325.151 Td /F4 8.7 Tf [(8)] TJ ET 0.271 0.267 0.267 rg BT 335.620 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 338.029 325.151 Td /F4 8.7 Tf [(9)] TJ ET 0.271 0.267 0.267 rg BT 342.848 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 345.257 325.151 Td /F4 8.7 Tf [(10)] TJ ET 0.271 0.267 0.267 rg BT 354.894 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 357.304 325.151 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 366.941 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 369.350 325.151 Td /F4 8.7 Tf [(12)] TJ ET 0.271 0.267 0.267 rg BT 378.988 327.532 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 381.397 325.151 Td /F4 8.7 Tf [(29)] TJ ET 0.271 0.267 0.267 rg BT 391.034 323.644 Td /F1 9.8 Tf [( messages and Google search queries )] TJ ET 0.267 0.267 0.267 rg BT 560.655 325.151 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 570.292 323.644 Td /F1 9.8 Tf [( )] TJ ET BT 26.250 311.739 Td /F1 9.8 Tf [(in order to detect influenza rates in real time \(i.e., influenza surveillance )] TJ ET 0.267 0.267 0.267 rg BT 336.212 313.247 Td /F4 8.7 Tf [(14)] TJ ET 0.271 0.267 0.267 rg BT 345.850 315.628 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 348.259 313.247 Td /F4 8.7 Tf [(15)] TJ ET 0.271 0.267 0.267 rg BT 357.896 315.628 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 360.306 313.247 Td /F4 8.7 Tf [(16)] TJ ET 0.271 0.267 0.267 rg BT 369.943 311.739 Td /F1 9.8 Tf [(\). Although Google Flu Trends \(GFT\) has )] TJ ET BT 26.250 299.835 Td /F1 9.8 Tf [(demonstrated some forecast accuracy, )] TJ ET 0.267 0.267 0.267 rg BT 196.943 301.342 Td /F4 8.7 Tf [(2)] TJ ET 0.271 0.267 0.267 rg BT 201.762 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 204.171 301.342 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 213.809 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 216.218 301.342 Td /F4 8.7 Tf [(18)] TJ ET 0.271 0.267 0.267 rg BT 225.855 303.723 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 228.265 301.342 Td /F4 8.7 Tf [(19)] TJ ET 0.271 0.267 0.267 rg BT 237.902 299.835 Td /F1 9.8 Tf [( it has recently been criticized because of its sensitivity to media reports, the )] TJ ET BT 26.250 287.930 Td /F1 9.8 Tf [(lack of transparency behind GFT data, and the infrequency with which GFT models are updated. )] TJ ET 0.267 0.267 0.267 rg BT 444.613 289.437 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 454.250 291.818 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 456.659 289.437 Td /F4 8.7 Tf [(20)] TJ ET 0.271 0.267 0.267 rg BT 466.297 287.930 Td /F1 9.8 Tf [( In contrast, the )] TJ ET BT 26.250 276.025 Td /F1 9.8 Tf [(forecasting potential of open social media, and Twitter in particular, remains largely untested.)] TJ ET BT 26.250 256.620 Td /F1 9.8 Tf [(In this paper, we demonstrate that influenza surveillance signals from Twitter significantly improve influenza forecasting. We use )] TJ ET BT 26.250 244.716 Td /F1 9.8 Tf [(freely available Twitter data and methods that are insensitive to influence from the mainstream media. )] TJ ET 0.267 0.267 0.267 rg BT 467.350 246.223 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 476.987 248.604 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 479.396 246.223 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 489.034 244.716 Td /F1 9.8 Tf [( We are the first to )] TJ ET BT 26.250 232.811 Td /F1 9.8 Tf [(perform an explicit forecast of influenza prevalence rates weeks into the future using social media data, and the first to compare )] TJ ET BT 26.250 220.906 Td /F1 9.8 Tf [(social media to GFT.)] TJ ET 0.267 0.267 0.267 rg BT 116.194 222.413 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 125.831 224.794 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 128.240 222.413 Td /F4 8.7 Tf [(30)] TJ ET 0.271 0.267 0.267 rg BT 26.250 201.501 Td /F1 9.8 Tf [(Our findings indicate that Twitter data are both more accessible, and provide better forecasts, when compared to GFT data. )] TJ ET 0.267 0.267 0.267 rg BT 560.579 203.009 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 26.250 189.597 Td /F1 9.8 Tf [(This is an important validation of social media data sources for influenza surveillance and forecasting.)] TJ ET BT 26.250 170.192 Td /F1 9.8 Tf [(Our analysis is also the first to use historically accurate surveillance data for the United States. Prior work has relied upon )] TJ ET BT 26.250 158.287 Td /F1 9.8 Tf [(amended data that were not available at the time the forecast was required. These data, consisting of Outpatient Influenza-like )] TJ ET BT 26.250 146.382 Td /F1 9.8 Tf [(Illness Surveillance Network \(ILINet\) reports from the U.S. Centers for Disease Control and Prevention, are the gold standard )] TJ ET BT 26.250 134.478 Td /F1 9.8 Tf [(for United States influenza surveillance. ILINet data are published with a one-week lag, though some have interpreted the lag as )] TJ ET BT 26.250 122.573 Td /F1 9.8 Tf [(being two weeks, )] TJ ET 0.267 0.267 0.267 rg BT 103.743 124.080 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 113.380 122.573 Td /F1 9.8 Tf [( the difference depending the dates used in determining the lag \(first day of the reported week vs. the last.\) )] TJ ET BT 26.250 110.668 Td /F1 9.8 Tf [(Importantly, the numbers initially released in CDC reports are subject to future revisions as data from additional ILINet sentinel )] TJ ET BT 26.250 98.763 Td /F1 9.8 Tf [(sites arrive. Retrospective analyses of ILINet data generally rely upon the final numbers released by the CDC; not the data )] TJ ET BT 26.250 86.859 Td /F1 9.8 Tf [(initially available. The degree to which updates to ILINet data might impact forecast accuracy has not been previously )] TJ ET BT 26.250 74.954 Td /F1 9.8 Tf [(considered. \(The effect of revisions for forecasting in Latin America was recently examined in )] TJ ET 0.267 0.267 0.267 rg BT 429.949 76.461 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 439.586 74.954 Td /F1 9.8 Tf [( .\) Our results demonstrate that )] TJ ET BT 26.250 63.049 Td /F1 9.8 Tf [(these revisions make a significant difference in forecasting efficacy, further highlighting the benefits of using real-time social )] TJ ET BT 26.250 51.144 Td /F1 9.8 Tf [(media data such as Twitter.)] TJ ET Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(1)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj 8 0 obj << /Type /Font /Subtype /Type1 /Name /F1 /BaseFont /Helvetica /Encoding /WinAnsiEncoding >> endobj 9 0 obj << /Type /Font /Subtype /Type1 /Name /F2 /BaseFont /Times-Bold /Encoding /WinAnsiEncoding >> endobj 10 0 obj << /Type /Font /Subtype /Type1 /Name /F3 /BaseFont /Times-Italic /Encoding /WinAnsiEncoding >> endobj 11 0 obj << /Type /Font /Subtype /Type1 /Name /F4 /BaseFont /Helvetica-Bold /Encoding /WinAnsiEncoding >> endobj 12 0 obj << /Type /XObject /Subtype /Image /Width 500 /Height 52 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 500 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 144>> stream x1 0 'ݲ؎"e{dzAdzAdzAdzAdzAdzAdzAdzAdzAdzAdzAdzAtlM0\ endstream endobj 13 0 obj << /Type /XObject /Subtype /Image /Width 500 /Height 52 /SMask 12 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 500 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 9410>> stream xy\SGFIʎ ^dZq(Z}֥Z *EZwQA@A,!}oB S{1s,wb̜9sF" xʝP r' !;W$O83OOYfjSR$ÆL&t[[[MtRss7_IQQQuuw7D#˖-C%ׯ_YX "''[=|nnnw+,sߙ r\@,bS 4NBݭ @ť{ԨQG <666b2l```cccbbbnnnllagQM(ʪn!H:::l6a2 Jbnnnff5 Ac```dddnnf/E&陘 yx-˕JT*t\.\.wpp2d訪bX4 >Ѐn/-͛7d2b߅}*RiaaaccAPKK srr WϞ=d..."B)  eI$v2lhhhooOWl0LLL>EswDD:wx\.H$ݻweeej S{Vh/_ĉ'Oj#JҥKZ6cW^PZUU̙3YNXX˗%涽Swvv._, lddf333\gϞ ݽ lٲBz/^pqq hԨQAAA&&&ȷ~ ]vW^eXzzzN¾͆aΝ;ɾUxxxpp0xw^bbbLbdd`0pS -V(ϟ?2e ^r+]vxvv6X2֭[сk goQˬYX,։'|GXvٵ /TKղ**xr?x Je0+VyϞ=|P( Ο?)Ǐ))))d2966.X &#Fdgg#"~Wkkkݻd6liӐ?veyz^^A`&p8 ݻT*Dk׮ h.hh" W^! @6wy=~`ȑT*U"WWW?~|@rʔ)/_>u,6om4>p8*Rk3g 2 0 k9Ztww/^ĉjD"oy<^JJ Br}nݺe92^x1j(mmm/^diiicc%K @]]]`:thtt:::ЧJ!P(]]]B4j BGGG ,]d~gmh4JB&xPgݻ2$ Js\CCC͛7rHj[[ۃ$%%_obbbVVqjjjrr#Gvcʕ(IOOGSF BDGGz@!Huuu:::w hرL>B充eee,AÇ/a AdɒǏ{zz2 l( .,,,߾ɓ'GEEM0/56ħOZYY}7oLLL @D"X,A LMMdC+""aXGGJjgplO?z>LMM4P(b=U6n8m4Y`oozO2l6F6 UMMM ,ƍ50 d2MMM=/-++=z4At:t={z{{̆ pG1L/_...UUU8 }}}ׯ hʕ)((@M{܁<~>**d999Ν(X'? *IMMuwwǥy?r<--mȍ7H7n̛7/((h̙Ǐrׯ!ILLtuu^pϒ%K꼽gϞ:(JBI$Һu~zAA@ (,,lܸqBpݨV*ܼyw֭h rxF[[L;;;^j7TϜ9sD777WWZ:͟=%%˗VBrAA QQQ[P(ZZZ F|~49Iwwwii)HJdj>xJ=:;;W~4442dL&wM6q[ 6y?dXUA߿ӧk֬qppPksεkv5 аf???//'ܺu g}JHH$''۷xGFF駾'OtO```||qWq-X7888!!gq3foPPoDDDuu5-z{{.] NW~:L.sN].?~|cǎ>}[# 0x9o???R?p__i˗;;;9ΪU꫱cN0!//ʕ+4 y X䳳]]]5nnnᩩj___rrD3JRpr B:ׯqBBN@ w@ڮB__߼;v,_`aH$MIiCCiӦİlR9siӦq͘1qآ^~mbb*°|D" 9s̙3cccq F>}o,^' wMJJNOO'Hk֬y/+ahvsڵ^ZtZϟBdhh ֭[~3IsGfX1WFյ?F[zJOOڛ^ZZI$^Y_hh;]!p-?q℮#GДOۃ=(˖-ùXbҤIΜ9YYYao@#55A^{{{l䉝;wzyyjQ*T*֭[G}ׯ]]]ЧdΜ9^^^P(ohh.w}}}]]]lْkPs*T*{{{E"Q{{{eeeIIɃ8Ξ={bcc!ߠ ssw6*]?p  0 o bܹ, }JJj]u߾}= yϙ3租~•w~昐PYYaÆ nK :Rd2N.`yC%$$x;v ӧO322lv~~>ֻD"Y[[߸qŋ#Gdj|Wt:gp>}ZZZ*HΝ+ˁH$244trrXhvd2J;;;ϛ7oMTB.WVVj(d2W^Jm۶ݺu+..HTO:UTTZEЏGٸq#A׮]/ǏoooAP(TОٳg/#G>uuuRT__?>>{nkk >ɓl6{߾}oߎ1b>366uww}ݻw---[ yfNNJMNN#H8D"yyyڵ-99~PǏh4kkkRԤ ΝYZRRr1iaaaaa!C***T hzPSv(RYYR ##`XXXy]H˗/[[[wX ɽUUU, (lᩩ߲e D" Lxb/knnm۶M67nKF6p[&A.\1117opAÇ!B&/?Iqq1 b8<<ԩSQ;\./5ס.<<`0ݗ-[_yzzbZN83F"x& Qc O&LښMLL>qܹ_~Y͛WYY3w\ܜ:wܚE^ۿpp ߺu+ֈzI30 :LLLFyvէT*NR8+J5aɓ'gddtuua0 :i$ EH?GGGD E L޸q9s޽+ݷ=zs<<>>z*==Hnڴ  ḙ; B4{CC֦%III-A56L:i~wzac!r]]]˗/GSZZZlmmߏ+Sxˠk׮7o~ƭvݟ-V3$$$** M9zǏ[[[Ç v}i{{{tt4344T[Vwܹl B4r@L&GEEsϐ6hҤI87vX T5?z[BWWWFFf{O>ĉ]]]jSYYtRmN.\PuxPEΝ3118qΝ; Ξ=pBJ"lݺUGGg͚57o<{]II 333׮]{񴴴aÆy{{cO@ÇS_}gSSӴcǎedd9se<}4>QΜ9cddСCCGEDD 6Ν;`ydnnnTTɪUٷSҽ{S!++.j:N%5ѡ:CBBԆ-//?rȔ)SpGQQQǎC/۷o^Zhbbbb58}c2OnG>r䈪k˖-666&MUOoܸaiiioolggK@hh(.\*&$$DFFΞ=k``8bGGGwww4o\\\RR.7ou>?}t0*ʲH333 qр ` x<^ss3koow)XZZh@XrU]C.ÇP|{0 [YYi<)okk#CBmhhhiib&9tPKKK+++wJvttԈb*jbbҟ eeed2 KKKX,^AH$#DǍ_p8htvv֢-H$'''_]] %˛( p fff/kkkmmѣ݂檪*Baeeh ^cvyќ9sRI+A} W ^B$''g~~~P(S(Grjd2hvP2d2 Ǔ'O@wPh``a" vO0 N: -V/^O+V(]oD># ̀5AaeAd,aFmyر۶mcXfff֭CCLP(J" xW~? ; >@N@@@'ZG endstream endobj 14 0 obj << /Type /Annot /Subtype /Link /A 15 0 R /Border [0 0 0] /H /I /Rect [ 15.0000 716.1516 373.7010 736.9416 ] >> endobj 15 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/article/twitter-improves-influenza-forecasting/) >> endobj 16 0 obj << /Type /Annot /Subtype /Link /A 17 0 R /Border [0 0 0] /H /I /Rect [ 93.2632 699.1581 158.5200 708.8106 ] >> endobj 17 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/article_category/research-article/) >> endobj 18 0 obj << /Type /Annot /Subtype /Link /A 19 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 687.2933 92.3550 697.2140 ] >> endobj 19 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/mpaulcs-jhu-edu/) >> endobj 20 0 obj << /Type /Annot /Subtype /Link /A 21 0 R /Border [0 0 0] /H /I /Rect [ 97.7760 687.2933 153.5753 697.2140 ] >> endobj 21 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/mdredze/) >> endobj 22 0 obj << /Type /Annot /Subtype /Link /A 23 0 R /Border [0 0 0] /H /I /Rect [ 158.9963 687.2933 241.8907 697.2140 ] >> endobj 23 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/david-broniatowskigmail-com/) >> endobj 24 0 obj << /Type /Annot /Subtype /Link /A 25 0 R /Border [0 0 0] /H /I /Rect [ 426.1658 379.4687 430.9844 388.2870 ] >> endobj 25 0 obj << /Type /Action >> endobj 26 0 obj << /Type /Annot /Subtype /Link /A 27 0 R /Border [0 0 0] /H /I /Rect [ 433.3938 379.4687 438.2124 388.2870 ] >> endobj 27 0 obj << /Type /Action >> endobj 28 0 obj << /Type /Annot /Subtype /Link /A 29 0 R /Border [0 0 0] /H /I /Rect [ 454.3433 355.6592 459.1619 364.4775 ] >> endobj 29 0 obj << /Type /Action >> endobj 30 0 obj << /Type /Annot /Subtype /Link /A 31 0 R /Border [0 0 0] /H /I /Rect [ 461.5713 355.6592 466.3899 364.4775 ] >> endobj 31 0 obj << /Type /Action >> endobj 32 0 obj << /Type /Annot /Subtype /Link /A 33 0 R /Border [0 0 0] /H /I /Rect [ 468.7993 355.6592 473.6179 364.4775 ] >> endobj 33 0 obj << /Type /Action >> endobj 34 0 obj << /Type /Annot /Subtype /Link /A 35 0 R /Border [0 0 0] /H /I /Rect [ 476.0273 355.6592 480.8459 364.4775 ] >> endobj 35 0 obj << /Type /Action >> endobj 36 0 obj << /Type /Annot /Subtype /Link /A 37 0 R /Border [0 0 0] /H /I /Rect [ 309.1170 324.3497 313.9357 333.1680 ] >> endobj 37 0 obj << /Type /Action >> endobj 38 0 obj << /Type /Annot /Subtype /Link /A 39 0 R /Border [0 0 0] /H /I /Rect [ 316.3450 324.3497 321.1637 333.1680 ] >> endobj 39 0 obj << /Type /Action >> endobj 40 0 obj << /Type /Annot /Subtype /Link /A 41 0 R /Border [0 0 0] /H /I /Rect [ 323.5730 324.3497 328.3917 333.1680 ] >> endobj 41 0 obj << /Type /Action >> endobj 42 0 obj << /Type /Annot /Subtype /Link /A 43 0 R /Border [0 0 0] /H /I /Rect [ 330.8010 324.3497 335.6197 333.1680 ] >> endobj 43 0 obj << /Type /Action >> endobj 44 0 obj << /Type /Annot /Subtype /Link /A 45 0 R /Border [0 0 0] /H /I /Rect [ 338.0290 324.3497 342.8477 333.1680 ] >> endobj 45 0 obj << /Type /Action >> endobj 46 0 obj << /Type /Annot /Subtype /Link /A 47 0 R /Border [0 0 0] /H /I /Rect [ 345.2570 324.3497 354.8943 333.1680 ] >> endobj 47 0 obj << /Type /Action >> endobj 48 0 obj << /Type /Annot /Subtype /Link /A 49 0 R /Border [0 0 0] /H /I /Rect [ 357.3037 324.3497 366.9410 333.1680 ] >> endobj 49 0 obj << /Type /Action >> endobj 50 0 obj << /Type /Annot /Subtype /Link /A 51 0 R /Border [0 0 0] /H /I /Rect [ 369.3503 324.3497 378.9877 333.1680 ] >> endobj 51 0 obj << /Type /Action >> endobj 52 0 obj << /Type /Annot /Subtype /Link /A 53 0 R /Border [0 0 0] /H /I /Rect [ 381.3970 324.3497 391.0343 333.1680 ] >> endobj 53 0 obj << /Type /Action >> endobj 54 0 obj << /Type /Annot /Subtype /Link /A 55 0 R /Border [0 0 0] /H /I /Rect [ 560.6551 324.3497 570.2924 333.1680 ] >> endobj 55 0 obj << /Type /Action >> endobj 56 0 obj << /Type /Annot /Subtype /Link /A 57 0 R /Border [0 0 0] /H /I /Rect [ 336.2122 312.4449 345.8496 321.2633 ] >> endobj 57 0 obj << /Type /Action >> endobj 58 0 obj << /Type /Annot /Subtype /Link /A 59 0 R /Border [0 0 0] /H /I /Rect [ 348.2589 312.4449 357.8963 321.2633 ] >> endobj 59 0 obj << /Type /Action >> endobj 60 0 obj << /Type /Annot /Subtype /Link /A 61 0 R /Border [0 0 0] /H /I /Rect [ 360.3056 312.4449 369.9429 321.2633 ] >> endobj 61 0 obj << /Type /Action >> endobj 62 0 obj << /Type /Annot /Subtype /Link /A 63 0 R /Border [0 0 0] /H /I /Rect [ 196.9433 300.5402 201.7619 309.3585 ] >> endobj 63 0 obj << /Type /Action >> endobj 64 0 obj << /Type /Annot /Subtype /Link /A 65 0 R /Border [0 0 0] /H /I /Rect [ 204.1713 300.5402 213.8086 309.3585 ] >> endobj 65 0 obj << /Type /Action >> endobj 66 0 obj << /Type /Annot /Subtype /Link /A 67 0 R /Border [0 0 0] /H /I /Rect [ 216.2179 300.5402 225.8553 309.3585 ] >> endobj 67 0 obj << /Type /Action >> endobj 68 0 obj << /Type /Annot /Subtype /Link /A 69 0 R /Border [0 0 0] /H /I /Rect [ 228.2646 300.5402 237.9019 309.3585 ] >> endobj 69 0 obj << /Type /Action >> endobj 70 0 obj << /Type /Annot /Subtype /Link /A 71 0 R /Border [0 0 0] /H /I /Rect [ 444.6128 288.6354 454.2501 297.4538 ] >> endobj 71 0 obj << /Type /Action >> endobj 72 0 obj << /Type /Annot /Subtype /Link /A 73 0 R /Border [0 0 0] /H /I /Rect [ 456.6594 288.6354 466.2968 297.4538 ] >> endobj 73 0 obj << /Type /Action >> endobj 74 0 obj << /Type /Annot /Subtype /Link /A 75 0 R /Border [0 0 0] /H /I /Rect [ 467.3497 245.4212 476.9871 254.2395 ] >> endobj 75 0 obj << /Type /Action >> endobj 76 0 obj << /Type /Annot /Subtype /Link /A 77 0 R /Border [0 0 0] /H /I /Rect [ 479.3964 245.4212 489.0337 254.2395 ] >> endobj 77 0 obj << /Type /Action >> endobj 78 0 obj << /Type /Annot /Subtype /Link /A 79 0 R /Border [0 0 0] /H /I /Rect [ 116.1937 221.6117 125.8311 230.4300 ] >> endobj 79 0 obj << /Type /Action >> endobj 80 0 obj << /Type /Annot /Subtype /Link /A 81 0 R /Border [0 0 0] /H /I /Rect [ 128.2404 221.6117 137.8778 230.4300 ] >> endobj 81 0 obj << /Type /Action >> endobj 82 0 obj << /Type /Annot /Subtype /Link /A 83 0 R /Border [0 0 0] /H /I /Rect [ 560.5793 202.2069 570.2166 211.0253 ] >> endobj 83 0 obj << /Type /Action >> endobj 84 0 obj << /Type /Annot /Subtype /Link /A 85 0 R /Border [0 0 0] /H /I /Rect [ 103.7430 123.2784 113.3803 132.0968 ] >> endobj 85 0 obj << /Type /Action >> endobj 86 0 obj << /Type /Annot /Subtype /Link /A 87 0 R /Border [0 0 0] /H /I /Rect [ 429.9488 75.6594 439.5861 84.4778 ] >> endobj 87 0 obj << /Type /Action >> endobj 88 0 obj << /Type /Annot /Subtype /Link /A 89 0 R /Border [0 0 0] /H /I /Rect [ 15.0000 716.1516 373.7010 736.9416 ] >> endobj 89 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/article/twitter-improves-influenza-forecasting/) >> endobj 90 0 obj << /Type /Annot /Subtype /Link /A 91 0 R /Border [0 0 0] /H /I /Rect [ 93.2632 699.1581 158.5200 708.8106 ] >> endobj 91 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/article_category/research-article/) >> endobj 92 0 obj << /Type /Annot /Subtype /Link /A 93 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 687.2933 92.3550 697.2140 ] >> endobj 93 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/mpaulcs-jhu-edu/) >> endobj 94 0 obj << /Type /Annot /Subtype /Link /A 95 0 R /Border [0 0 0] /H /I /Rect [ 97.7760 687.2933 153.5753 697.2140 ] >> endobj 95 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/mdredze/) >> endobj 96 0 obj << /Type /Annot /Subtype /Link /A 97 0 R /Border [0 0 0] /H /I /Rect [ 158.9963 687.2933 241.8907 697.2140 ] >> endobj 97 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/david-broniatowskigmail-com/) >> endobj 98 0 obj << /Type /Annot /Subtype /Link /A 99 0 R /Border [0 0 0] /H /I /Rect [ 426.1658 379.4687 430.9844 388.2870 ] >> endobj 99 0 obj << /Type /Action >> endobj 100 0 obj << /Type /Annot /Subtype /Link /A 101 0 R /Border [0 0 0] /H /I /Rect [ 433.3938 379.4687 438.2124 388.2870 ] >> endobj 101 0 obj << /Type /Action >> endobj 102 0 obj << /Type /Annot /Subtype /Link /A 103 0 R /Border [0 0 0] /H /I /Rect [ 454.3433 355.6592 459.1619 364.4775 ] >> endobj 103 0 obj << /Type /Action >> endobj 104 0 obj << /Type /Annot /Subtype /Link /A 105 0 R /Border [0 0 0] /H /I /Rect [ 461.5713 355.6592 466.3899 364.4775 ] >> endobj 105 0 obj << /Type /Action >> endobj 106 0 obj << /Type /Annot /Subtype /Link /A 107 0 R /Border [0 0 0] /H /I /Rect [ 468.7993 355.6592 473.6179 364.4775 ] >> endobj 107 0 obj << /Type /Action >> endobj 108 0 obj << /Type /Annot /Subtype /Link /A 109 0 R /Border [0 0 0] /H /I /Rect [ 476.0273 355.6592 480.8459 364.4775 ] >> endobj 109 0 obj << /Type /Action >> endobj 110 0 obj << /Type /Annot /Subtype /Link /A 111 0 R /Border [0 0 0] /H /I /Rect [ 309.1170 324.3497 313.9357 333.1680 ] >> endobj 111 0 obj << /Type /Action >> endobj 112 0 obj << /Type /Annot /Subtype /Link /A 113 0 R /Border [0 0 0] /H /I /Rect [ 316.3450 324.3497 321.1637 333.1680 ] >> endobj 113 0 obj << /Type /Action >> endobj 114 0 obj << /Type /Annot /Subtype /Link /A 115 0 R /Border [0 0 0] /H /I /Rect [ 323.5730 324.3497 328.3917 333.1680 ] >> endobj 115 0 obj << /Type /Action >> endobj 116 0 obj << /Type /Annot /Subtype /Link /A 117 0 R /Border [0 0 0] /H /I /Rect [ 330.8010 324.3497 335.6197 333.1680 ] >> endobj 117 0 obj << /Type /Action >> endobj 118 0 obj << /Type /Annot /Subtype /Link /A 119 0 R /Border [0 0 0] /H /I /Rect [ 338.0290 324.3497 342.8477 333.1680 ] >> endobj 119 0 obj << /Type /Action >> endobj 120 0 obj << /Type /Annot /Subtype /Link /A 121 0 R /Border [0 0 0] /H /I /Rect [ 345.2570 324.3497 354.8943 333.1680 ] >> endobj 121 0 obj << /Type /Action >> endobj 122 0 obj << /Type /Annot /Subtype /Link /A 123 0 R /Border [0 0 0] /H /I /Rect [ 357.3037 324.3497 366.9410 333.1680 ] >> endobj 123 0 obj << /Type /Action >> endobj 124 0 obj << /Type /Annot /Subtype /Link /A 125 0 R /Border [0 0 0] /H /I /Rect [ 369.3503 324.3497 378.9877 333.1680 ] >> endobj 125 0 obj << /Type /Action >> endobj 126 0 obj << /Type /Annot /Subtype /Link /A 127 0 R /Border [0 0 0] /H /I /Rect [ 381.3970 324.3497 391.0343 333.1680 ] >> endobj 127 0 obj << /Type /Action >> endobj 128 0 obj << /Type /Annot /Subtype /Link /A 129 0 R /Border [0 0 0] /H /I /Rect [ 560.6551 324.3497 570.2924 333.1680 ] >> endobj 129 0 obj << /Type /Action >> endobj 130 0 obj << /Type /Annot /Subtype /Link /A 131 0 R /Border [0 0 0] /H /I /Rect [ 336.2122 312.4449 345.8496 321.2633 ] >> endobj 131 0 obj << /Type /Action >> endobj 132 0 obj << /Type /Annot /Subtype /Link /A 133 0 R /Border [0 0 0] /H /I /Rect [ 348.2589 312.4449 357.8963 321.2633 ] >> endobj 133 0 obj << /Type /Action >> endobj 134 0 obj << /Type /Annot /Subtype /Link /A 135 0 R /Border [0 0 0] /H /I /Rect [ 360.3056 312.4449 369.9429 321.2633 ] >> endobj 135 0 obj << /Type /Action >> endobj 136 0 obj << /Type /Annot /Subtype /Link /A 137 0 R /Border [0 0 0] /H /I /Rect [ 196.9433 300.5402 201.7619 309.3585 ] >> endobj 137 0 obj << /Type /Action >> endobj 138 0 obj << /Type /Annot /Subtype /Link /A 139 0 R /Border [0 0 0] /H /I /Rect [ 204.1713 300.5402 213.8086 309.3585 ] >> endobj 139 0 obj << /Type /Action >> endobj 140 0 obj << /Type /Annot /Subtype /Link /A 141 0 R /Border [0 0 0] /H /I /Rect [ 216.2179 300.5402 225.8553 309.3585 ] >> endobj 141 0 obj << /Type /Action >> endobj 142 0 obj << /Type /Annot /Subtype /Link /A 143 0 R /Border [0 0 0] /H /I /Rect [ 228.2646 300.5402 237.9019 309.3585 ] >> endobj 143 0 obj << /Type /Action >> endobj 144 0 obj << /Type /Annot /Subtype /Link /A 145 0 R /Border [0 0 0] /H /I /Rect [ 444.6128 288.6354 454.2501 297.4538 ] >> endobj 145 0 obj << /Type /Action >> endobj 146 0 obj << /Type /Annot /Subtype /Link /A 147 0 R /Border [0 0 0] /H /I /Rect [ 456.6594 288.6354 466.2968 297.4538 ] >> endobj 147 0 obj << /Type /Action >> endobj 148 0 obj << /Type /Annot /Subtype /Link /A 149 0 R /Border [0 0 0] /H /I /Rect [ 467.3497 245.4212 476.9871 254.2395 ] >> endobj 149 0 obj << /Type /Action >> endobj 150 0 obj << /Type /Annot /Subtype /Link /A 151 0 R /Border [0 0 0] /H /I /Rect [ 479.3964 245.4212 489.0337 254.2395 ] >> endobj 151 0 obj << /Type /Action >> endobj 152 0 obj << /Type /Annot /Subtype /Link /A 153 0 R /Border [0 0 0] /H /I /Rect [ 116.1937 221.6117 125.8311 230.4300 ] >> endobj 153 0 obj << /Type /Action >> endobj 154 0 obj << /Type /Annot /Subtype /Link /A 155 0 R /Border [0 0 0] /H /I /Rect [ 128.2404 221.6117 137.8778 230.4300 ] >> endobj 155 0 obj << /Type /Action >> endobj 156 0 obj << /Type /Annot /Subtype /Link /A 157 0 R /Border [0 0 0] /H /I /Rect [ 560.5793 202.2069 570.2166 211.0253 ] >> endobj 157 0 obj << /Type /Action >> endobj 158 0 obj << /Type /Annot /Subtype /Link /A 159 0 R /Border [0 0 0] /H /I /Rect [ 103.7430 123.2784 113.3803 132.0968 ] >> endobj 159 0 obj << /Type /Action >> endobj 160 0 obj << /Type /Annot /Subtype /Link /A 161 0 R /Border [0 0 0] /H /I /Rect [ 429.9488 75.6594 439.5861 84.4778 ] >> endobj 161 0 obj << /Type /Action >> endobj 162 0 obj << /Type /Annot /Subtype /Link /A 163 0 R /Border [0 0 0] /H /I /Rect [ 15.0000 716.1516 373.7010 736.9416 ] >> endobj 163 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/article/twitter-improves-influenza-forecasting/) >> endobj 164 0 obj << /Type /Annot /Subtype /Link /A 165 0 R /Border [0 0 0] /H /I /Rect [ 93.2632 699.1581 158.5200 708.8106 ] >> endobj 165 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/article_category/research-article/) >> endobj 166 0 obj << /Type /Annot /Subtype /Link /A 167 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 687.2933 92.3550 697.2140 ] >> endobj 167 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/mpaulcs-jhu-edu/) >> endobj 168 0 obj << /Type /Annot /Subtype /Link /A 169 0 R /Border [0 0 0] /H /I /Rect [ 97.7760 687.2933 153.5753 697.2140 ] >> endobj 169 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/mdredze/) >> endobj 170 0 obj << /Type /Annot /Subtype /Link /A 171 0 R /Border [0 0 0] /H /I /Rect [ 158.9963 687.2933 241.8907 697.2140 ] >> endobj 171 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/author/david-broniatowskigmail-com/) >> endobj 172 0 obj << /Type /Annot /Subtype /Link /A 173 0 R /Border [0 0 0] /H /I /Rect [ 426.1658 379.4687 430.9844 388.2870 ] >> endobj 173 0 obj << /Type /Action >> endobj 174 0 obj << /Type /Annot /Subtype /Link /A 175 0 R /Border [0 0 0] /H /I /Rect [ 433.3938 379.4687 438.2124 388.2870 ] >> endobj 175 0 obj << /Type /Action >> endobj 176 0 obj << /Type /Annot /Subtype /Link /A 177 0 R /Border [0 0 0] /H /I /Rect [ 454.3433 355.6592 459.1619 364.4775 ] >> endobj 177 0 obj << /Type /Action >> endobj 178 0 obj << /Type /Annot /Subtype /Link /A 179 0 R /Border [0 0 0] /H /I /Rect [ 461.5713 355.6592 466.3899 364.4775 ] >> endobj 179 0 obj << /Type /Action >> endobj 180 0 obj << /Type /Annot /Subtype /Link /A 181 0 R /Border [0 0 0] /H /I /Rect [ 468.7993 355.6592 473.6179 364.4775 ] >> endobj 181 0 obj << /Type /Action >> endobj 182 0 obj << /Type /Annot /Subtype /Link /A 183 0 R /Border [0 0 0] /H /I /Rect [ 476.0273 355.6592 480.8459 364.4775 ] >> endobj 183 0 obj << /Type /Action >> endobj 184 0 obj << /Type /Annot /Subtype /Link /A 185 0 R /Border [0 0 0] /H /I /Rect [ 309.1170 324.3497 313.9357 333.1680 ] >> endobj 185 0 obj << /Type /Action >> endobj 186 0 obj << /Type /Annot /Subtype /Link /A 187 0 R /Border [0 0 0] /H /I /Rect [ 316.3450 324.3497 321.1637 333.1680 ] >> endobj 187 0 obj << /Type /Action >> endobj 188 0 obj << /Type /Annot /Subtype /Link /A 189 0 R /Border [0 0 0] /H /I /Rect [ 323.5730 324.3497 328.3917 333.1680 ] >> endobj 189 0 obj << /Type /Action >> endobj 190 0 obj << /Type /Annot /Subtype /Link /A 191 0 R /Border [0 0 0] /H /I /Rect [ 330.8010 324.3497 335.6197 333.1680 ] >> endobj 191 0 obj << /Type /Action >> endobj 192 0 obj << /Type /Annot /Subtype /Link /A 193 0 R /Border [0 0 0] /H /I /Rect [ 338.0290 324.3497 342.8477 333.1680 ] >> endobj 193 0 obj << /Type /Action >> endobj 194 0 obj << /Type /Annot /Subtype /Link /A 195 0 R /Border [0 0 0] /H /I /Rect [ 345.2570 324.3497 354.8943 333.1680 ] >> endobj 195 0 obj << /Type /Action >> endobj 196 0 obj << /Type /Annot /Subtype /Link /A 197 0 R /Border [0 0 0] /H /I /Rect [ 357.3037 324.3497 366.9410 333.1680 ] >> endobj 197 0 obj << /Type /Action >> endobj 198 0 obj << /Type /Annot /Subtype /Link /A 199 0 R /Border [0 0 0] /H /I /Rect [ 369.3503 324.3497 378.9877 333.1680 ] >> endobj 199 0 obj << /Type /Action >> endobj 200 0 obj << /Type /Annot /Subtype /Link /A 201 0 R /Border [0 0 0] /H /I /Rect [ 381.3970 324.3497 391.0343 333.1680 ] >> endobj 201 0 obj << /Type /Action >> endobj 202 0 obj << /Type /Annot /Subtype /Link /A 203 0 R /Border [0 0 0] /H /I /Rect [ 560.6551 324.3497 570.2924 333.1680 ] >> endobj 203 0 obj << /Type /Action >> endobj 204 0 obj << /Type /Annot /Subtype /Link /A 205 0 R /Border [0 0 0] /H /I /Rect [ 336.2122 312.4449 345.8496 321.2633 ] >> endobj 205 0 obj << /Type /Action >> endobj 206 0 obj << /Type /Annot /Subtype /Link /A 207 0 R /Border [0 0 0] /H /I /Rect [ 348.2589 312.4449 357.8963 321.2633 ] >> endobj 207 0 obj << /Type /Action >> endobj 208 0 obj << /Type /Annot /Subtype /Link /A 209 0 R /Border [0 0 0] /H /I /Rect [ 360.3056 312.4449 369.9429 321.2633 ] >> endobj 209 0 obj << /Type /Action >> endobj 210 0 obj << /Type /Annot /Subtype /Link /A 211 0 R /Border [0 0 0] /H /I /Rect [ 196.9433 300.5402 201.7619 309.3585 ] >> endobj 211 0 obj << /Type /Action >> endobj 212 0 obj << /Type /Annot /Subtype /Link /A 213 0 R /Border [0 0 0] /H /I /Rect [ 204.1713 300.5402 213.8086 309.3585 ] >> endobj 213 0 obj << /Type /Action >> endobj 214 0 obj << /Type /Annot /Subtype /Link /A 215 0 R /Border [0 0 0] /H /I /Rect [ 216.2179 300.5402 225.8553 309.3585 ] >> endobj 215 0 obj << /Type /Action >> endobj 216 0 obj << /Type /Annot /Subtype /Link /A 217 0 R /Border [0 0 0] /H /I /Rect [ 228.2646 300.5402 237.9019 309.3585 ] >> endobj 217 0 obj << /Type /Action >> endobj 218 0 obj << /Type /Annot /Subtype /Link /A 219 0 R /Border [0 0 0] /H /I /Rect [ 444.6128 288.6354 454.2501 297.4538 ] >> endobj 219 0 obj << /Type /Action >> endobj 220 0 obj << /Type /Annot /Subtype /Link /A 221 0 R /Border [0 0 0] /H /I /Rect [ 456.6594 288.6354 466.2968 297.4538 ] >> endobj 221 0 obj << /Type /Action >> endobj 222 0 obj << /Type /Annot /Subtype /Link /A 223 0 R /Border [0 0 0] /H /I /Rect [ 467.3497 245.4212 476.9871 254.2395 ] >> endobj 223 0 obj << /Type /Action >> endobj 224 0 obj << /Type /Annot /Subtype /Link /A 225 0 R /Border [0 0 0] /H /I /Rect [ 479.3964 245.4212 489.0337 254.2395 ] >> endobj 225 0 obj << /Type /Action >> endobj 226 0 obj << /Type /Annot /Subtype /Link /A 227 0 R /Border [0 0 0] /H /I /Rect [ 116.1937 221.6117 125.8311 230.4300 ] >> endobj 227 0 obj << /Type /Action >> endobj 228 0 obj << /Type /Annot /Subtype /Link /A 229 0 R /Border [0 0 0] /H /I /Rect [ 128.2404 221.6117 137.8778 230.4300 ] >> endobj 229 0 obj << /Type /Action >> endobj 230 0 obj << /Type /Annot /Subtype /Link /A 231 0 R /Border [0 0 0] /H /I /Rect [ 560.5793 202.2069 570.2166 211.0253 ] >> endobj 231 0 obj << /Type /Action >> endobj 232 0 obj << /Type /Annot /Subtype /Link /A 233 0 R /Border [0 0 0] /H /I /Rect [ 103.7430 123.2784 113.3803 132.0968 ] >> endobj 233 0 obj << /Type /Action >> endobj 234 0 obj << /Type /Annot /Subtype /Link /A 235 0 R /Border [0 0 0] /H /I /Rect [ 429.9488 75.6594 439.5861 84.4778 ] >> endobj 235 0 obj << /Type /Action >> endobj 236 0 obj << /Type /Page /Parent 3 0 R /Annots [ 239 0 R 241 0 R 243 0 R 245 0 R 247 0 R 249 0 R 251 0 R 253 0 R 255 0 R 259 0 R 271 0 R 273 0 R 275 0 R 277 0 R 279 0 R 281 0 R 283 0 R 285 0 R 287 0 R 291 0 R 303 0 R 305 0 R 307 0 R 309 0 R 311 0 R 313 0 R 315 0 R 317 0 R 319 0 R 323 0 R ] /Contents 237 0 R >> endobj 237 0 obj << /Length 26668 >> stream 0.271 0.267 0.267 rg q 15.000 45.598 577.500 731.402 re W n 0.271 0.267 0.267 rg BT 26.250 750.278 Td /F4 12.0 Tf [(Methods)] TJ ET BT 26.250 730.324 Td /F4 9.8 Tf [(ILINet)] TJ ET BT 26.250 710.919 Td /F1 9.8 Tf [(Our baseline surveillance data and gold standard for predictions are based on CDCs ILINet. Historical and current ILINet data )] TJ ET BT 26.250 699.015 Td /F1 9.8 Tf [(\(at the national, HHS, and Census-division regional levels\) are available from: )] TJ ET 0.267 0.267 0.267 rg BT 26.250 687.110 Td /F1 9.8 Tf [(http://gis.cdc.gov/grasp/fluview/fluportaldashboard.html)] TJ ET 0.271 0.267 0.267 rg BT 263.604 687.110 Td /F1 9.8 Tf [(. Our analyses use the weighted version of this metric, which adjusts for )] TJ ET BT 26.250 675.205 Td /F1 9.8 Tf [(state population. Data are available starting with the 1997-1998 influenza season. Weekly tables released on week )] TJ ET BT 523.207 675.205 Td /F4 9.8 Tf [(W)] TJ ET BT 532.411 675.205 Td /F1 9.8 Tf [( of the )] TJ ET BT 562.227 675.205 Td /F4 9.8 Tf [(X-Y )] TJ ET BT 26.250 663.300 Td /F1 9.8 Tf [(season are available at the following URLs:)] TJ ET BT 26.250 643.896 Td /F1 9.8 Tf [(http://www.cdc.gov/flu/weekly/weeklyarchives)] TJ ET BT 222.391 643.896 Td /F4 9.8 Tf [(X-Y)] TJ ET BT 238.644 643.896 Td /F1 9.8 Tf [(/data/senAllregt)] TJ ET BT 306.387 643.896 Td /F4 9.8 Tf [(W)] TJ ET BT 315.591 643.896 Td /F1 9.8 Tf [(.htm.)] TJ ET BT 26.250 624.491 Td /F1 9.8 Tf [(In addition to final ILINet values, we downloaded ILINet data that were available at a particular time from these tables. Such )] TJ ET BT 26.250 612.586 Td /F1 9.8 Tf [(tables are available for all seasons beginning 2004-2005.)] TJ ET BT 26.250 593.181 Td /F4 9.8 Tf [(Twitter)] TJ ET BT 26.250 573.777 Td /F1 9.8 Tf [(We use the Twitter influenza surveillance system developed by Lamb et al. )] TJ ET 0.267 0.267 0.267 rg BT 351.383 575.284 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 361.021 577.665 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 363.430 575.284 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 373.067 573.777 Td /F1 9.8 Tf [( to produce weekly influenza rates, since it )] TJ ET BT 26.250 561.872 Td /F1 9.8 Tf [(achieves state of the art surveillance results for Twitter data. This algorithm identifies such messages using a cascade of logistic )] TJ ET BT 26.250 549.967 Td /F1 9.8 Tf [(regression classifiers that determine, first, if a message is about health; next, if it is about influenza; and finally, if it is about an )] TJ ET BT 26.250 538.062 Td /F1 9.8 Tf [(influenza infection \(rather than simply an awareness of the ongoing flu season\). The classifiers are trained purely on the )] TJ ET BT 26.250 526.158 Td /F1 9.8 Tf [(message content rather than on historical ILI data. We used messages from the United States as determined by the )] TJ ET BT 527.536 526.158 Td /F5 9.8 Tf [(Carmen)] TJ ET BT 26.250 514.253 Td /F1 9.8 Tf [(geolocation system. )] TJ ET 0.267 0.267 0.267 rg BT 115.121 515.760 Td /F4 8.7 Tf [(27)] TJ ET 0.271 0.267 0.267 rg BT 124.759 514.253 Td /F1 9.8 Tf [( We use the output of these models as features for the forecast model based on data from November 27, )] TJ ET BT 26.250 502.348 Td /F1 9.8 Tf [(2011 through April 5, 2014.)] TJ ET BT 26.250 482.943 Td /F4 9.8 Tf [(Google Flu Trends)] TJ ET BT 26.250 463.539 Td /F1 9.8 Tf [(Google Flu Trends is an influenza surveillance system that estimates current infection rates based on the volume of Google )] TJ ET BT 26.250 451.634 Td /F1 9.8 Tf [(searches for a select number of influenza related queries. )] TJ ET 0.267 0.267 0.267 rg BT 277.156 453.141 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 286.794 451.634 Td /F1 9.8 Tf [( GFT data are available from )] TJ ET 0.267 0.267 0.267 rg BT 26.250 439.729 Td /F1 9.8 Tf [(http://www.google.org/flutrends/us/data.txt)] TJ ET 0.271 0.267 0.267 rg BT 208.341 439.729 Td /F1 9.8 Tf [(. We collected all GFT estimates for the US and restrict our attention to the same time )] TJ ET BT 26.250 427.824 Td /F1 9.8 Tf [(interval as our Twitter data for a direct comparison. Following poor performance during the 2012-13 season, )] TJ ET 0.267 0.267 0.267 rg BT 493.363 429.332 Td /F4 8.7 Tf [(21)] TJ ET 0.271 0.267 0.267 rg BT 503.000 427.824 Td /F1 9.8 Tf [( the GFT model )] TJ ET BT 26.250 415.920 Td /F1 9.8 Tf [(was updated in August 2013. Therefore, the numbers beginning in this month are based on the latest model. Retrospective )] TJ ET BT 26.250 404.015 Td /F1 9.8 Tf [(estimates are also available for earlier data using the newest model, but because these estimates are based on a model trained )] TJ ET BT 26.250 392.110 Td /F1 9.8 Tf [(on the same data \(in-sample data\), they do not provide an accurate assessment of the models predictive abilities and we )] TJ ET BT 26.250 380.205 Td /F1 9.8 Tf [(therefore do not use them.)] TJ ET BT 26.250 360.801 Td /F4 9.8 Tf [(Model)] TJ ET BT 26.250 341.396 Td /F1 9.8 Tf [(When forecasting influenza rates, we used a basic linear autoregressive model for ease of comparison to previous work. )] TJ ET 0.267 0.267 0.267 rg BT 547.007 342.903 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 556.645 345.284 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 559.054 342.903 Td /F4 8.7 Tf [(31)] TJ ET 0.271 0.267 0.267 rg BT 26.250 329.491 Td /F1 9.8 Tf [(Our linear model took the form:)] TJ ET q 212.250 0 0 15.000 26.250 304.610 cm /I4 Do Q BT 26.250 287.586 Td /F1 9.8 Tf [(where y)] TJ ET BT 60.385 285.522 Td /F1 8.7 Tf [(w )] TJ ET BT 69.051 287.586 Td /F1 9.8 Tf [(denotes the ILI prevalence at week w and the values of a are the regression coefficients. When k=0, we are )] TJ ET BT 26.250 275.682 Td /F1 9.8 Tf [(nowcasting inferring the present influenza prevalence rate that the CDC will report in the following week. )] TJ ET 0.267 0.267 0.267 rg BT 492.271 277.189 Td /F4 8.7 Tf [(7)] TJ ET 0.271 0.267 0.267 rg BT 497.089 275.682 Td /F1 9.8 Tf [( When k > 0 we )] TJ ET BT 26.250 263.777 Td /F1 9.8 Tf [(are forecasting further into the future. The )] TJ ET q 8.250 0 0 7.500 208.887 265.801 cm /I6 Do Q BT 217.137 263.777 Td /F1 9.8 Tf [( parameters can be estimated using least-squares regression, where parameters )] TJ ET BT 26.250 251.872 Td /F1 9.8 Tf [(are estimated separately for different values of k.)] TJ ET BT 26.250 232.467 Td /F1 9.8 Tf [(We distinguished between the gold standard data that we are trying to estimate i.e., ILINet data that are no longer being )] TJ ET BT 26.250 220.563 Td /F1 9.8 Tf [(revised and the data that are actually available on week w-1 from the CDCs weekly reports. Whereas y denotes the weekly )] TJ ET BT 26.250 208.582 Td /F1 9.8 Tf [(value in the final gold standard report, )] TJ ET q 10.500 0 0 12.000 198.581 206.182 cm /I8 Do Q BT 209.081 208.582 Td /F1 9.8 Tf [( denotes the values that are published in the report that is most recent at the time of )] TJ ET BT 26.250 196.582 Td /F1 9.8 Tf [(the forecast. Our use of )] TJ ET q 10.500 0 0 12.000 130.848 194.182 cm /I10 Do Q BT 141.348 196.582 Td /F1 9.8 Tf [( in our model ensures that we more accurately reflect the expected performance of forecasts )] TJ ET BT 26.250 184.658 Td /F1 9.8 Tf [(produced using the most recent ILINet data. When forecasting, we trained our baseline model on three seasons of data, )] TJ ET BT 26.250 172.753 Td /F1 9.8 Tf [(beginning with the 2011-2012 season and evaluated it using cross-fold validation, where each season of data is a cross-)] TJ ET BT 26.250 160.848 Td /F1 9.8 Tf [(validation fold. \(We also trained with data beginning in 2004, yielding a negligible reduction in error: a decrease of 0.001.\))] TJ ET BT 26.250 141.444 Td /F1 9.8 Tf [(One advantage of both Twitter and GFT-based systems is that data are available for the current week, not just the previous )] TJ ET BT 26.250 129.539 Td /F1 9.8 Tf [(week. To augment the forecasting model, we include z)] TJ ET BT 261.440 127.475 Td /F1 8.7 Tf [(w)] TJ ET BT 267.697 129.539 Td /F1 9.8 Tf [(, the Web-based methods estimate for week w:)] TJ ET q 250.500 0 0 15.000 26.250 104.658 cm /I12 Do Q BT 26.250 87.634 Td /F1 9.8 Tf [(Similarly, we can utilize multiple Web-based information sources by adding additional terms for each z)] TJ ET BT 465.692 85.570 Td /F1 8.7 Tf [(w)] TJ ET BT 471.950 87.634 Td /F1 9.8 Tf [(.)] TJ ET BT 26.250 67.553 Td /F1 9.8 Tf [(For the purpose of evaluation, we also experimented with a model that uses )] TJ ET BT 355.741 67.553 Td /F5 9.8 Tf [(only)] TJ ET BT 373.623 67.553 Td /F1 9.8 Tf [( Web-based data, )] TJ ET q 54.750 0 0 12.750 453.290 65.003 cm /I14 Do Q BT 508.040 67.553 Td /F1 9.8 Tf [(, which fits the )] TJ ET BT 26.250 55.479 Td /F1 9.8 Tf [(web-based data to the corresponding ILINet values.)] TJ ET Q q 15.000 45.598 577.500 731.402 re W n 0.271 0.267 0.267 rg BT 26.250 750.278 Td /F4 12.0 Tf [(Methods)] TJ ET BT 26.250 730.324 Td /F4 9.8 Tf [(ILINet)] TJ ET BT 26.250 710.919 Td /F1 9.8 Tf [(Our baseline surveillance data and gold standard for predictions are based on CDCs ILINet. Historical and current ILINet data )] TJ ET BT 26.250 699.015 Td /F1 9.8 Tf [(\(at the national, HHS, and Census-division regional levels\) are available from: )] TJ ET 0.267 0.267 0.267 rg BT 26.250 687.110 Td /F1 9.8 Tf [(http://gis.cdc.gov/grasp/fluview/fluportaldashboard.html)] TJ ET 0.271 0.267 0.267 rg BT 263.604 687.110 Td /F1 9.8 Tf [(. Our analyses use the weighted version of this metric, which adjusts for )] TJ ET BT 26.250 675.205 Td /F1 9.8 Tf [(state population. Data are available starting with the 1997-1998 influenza season. Weekly tables released on week )] TJ ET BT 523.207 675.205 Td /F4 9.8 Tf [(W)] TJ ET BT 532.411 675.205 Td /F1 9.8 Tf [( of the )] TJ ET BT 562.227 675.205 Td /F4 9.8 Tf [(X-Y )] TJ ET BT 26.250 663.300 Td /F1 9.8 Tf [(season are available at the following URLs:)] TJ ET BT 26.250 643.896 Td /F1 9.8 Tf [(http://www.cdc.gov/flu/weekly/weeklyarchives)] TJ ET BT 222.391 643.896 Td /F4 9.8 Tf [(X-Y)] TJ ET BT 238.644 643.896 Td /F1 9.8 Tf [(/data/senAllregt)] TJ ET BT 306.387 643.896 Td /F4 9.8 Tf [(W)] TJ ET BT 315.591 643.896 Td /F1 9.8 Tf [(.htm.)] TJ ET BT 26.250 624.491 Td /F1 9.8 Tf [(In addition to final ILINet values, we downloaded ILINet data that were available at a particular time from these tables. Such )] TJ ET BT 26.250 612.586 Td /F1 9.8 Tf [(tables are available for all seasons beginning 2004-2005.)] TJ ET BT 26.250 593.181 Td /F4 9.8 Tf [(Twitter)] TJ ET BT 26.250 573.777 Td /F1 9.8 Tf [(We use the Twitter influenza surveillance system developed by Lamb et al. )] TJ ET 0.267 0.267 0.267 rg BT 351.383 575.284 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 361.021 577.665 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 363.430 575.284 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 373.067 573.777 Td /F1 9.8 Tf [( to produce weekly influenza rates, since it )] TJ ET BT 26.250 561.872 Td /F1 9.8 Tf [(achieves state of the art surveillance results for Twitter data. This algorithm identifies such messages using a cascade of logistic )] TJ ET BT 26.250 549.967 Td /F1 9.8 Tf [(regression classifiers that determine, first, if a message is about health; next, if it is about influenza; and finally, if it is about an )] TJ ET BT 26.250 538.062 Td /F1 9.8 Tf [(influenza infection \(rather than simply an awareness of the ongoing flu season\). The classifiers are trained purely on the )] TJ ET BT 26.250 526.158 Td /F1 9.8 Tf [(message content rather than on historical ILI data. We used messages from the United States as determined by the )] TJ ET BT 527.536 526.158 Td /F5 9.8 Tf [(Carmen)] TJ ET BT 26.250 514.253 Td /F1 9.8 Tf [(geolocation system. )] TJ ET 0.267 0.267 0.267 rg BT 115.121 515.760 Td /F4 8.7 Tf [(27)] TJ ET 0.271 0.267 0.267 rg BT 124.759 514.253 Td /F1 9.8 Tf [( We use the output of these models as features for the forecast model based on data from November 27, )] TJ ET BT 26.250 502.348 Td /F1 9.8 Tf [(2011 through April 5, 2014.)] TJ ET BT 26.250 482.943 Td /F4 9.8 Tf [(Google Flu Trends)] TJ ET BT 26.250 463.539 Td /F1 9.8 Tf [(Google Flu Trends is an influenza surveillance system that estimates current infection rates based on the volume of Google )] TJ ET BT 26.250 451.634 Td /F1 9.8 Tf [(searches for a select number of influenza related queries. )] TJ ET 0.267 0.267 0.267 rg BT 277.156 453.141 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 286.794 451.634 Td /F1 9.8 Tf [( GFT data are available from )] TJ ET 0.267 0.267 0.267 rg BT 26.250 439.729 Td /F1 9.8 Tf [(http://www.google.org/flutrends/us/data.txt)] TJ ET 0.271 0.267 0.267 rg BT 208.341 439.729 Td /F1 9.8 Tf [(. We collected all GFT estimates for the US and restrict our attention to the same time )] TJ ET BT 26.250 427.824 Td /F1 9.8 Tf [(interval as our Twitter data for a direct comparison. Following poor performance during the 2012-13 season, )] TJ ET 0.267 0.267 0.267 rg BT 493.363 429.332 Td /F4 8.7 Tf [(21)] TJ ET 0.271 0.267 0.267 rg BT 503.000 427.824 Td /F1 9.8 Tf [( the GFT model )] TJ ET BT 26.250 415.920 Td /F1 9.8 Tf [(was updated in August 2013. Therefore, the numbers beginning in this month are based on the latest model. Retrospective )] TJ ET BT 26.250 404.015 Td /F1 9.8 Tf [(estimates are also available for earlier data using the newest model, but because these estimates are based on a model trained )] TJ ET BT 26.250 392.110 Td /F1 9.8 Tf [(on the same data \(in-sample data\), they do not provide an accurate assessment of the models predictive abilities and we )] TJ ET BT 26.250 380.205 Td /F1 9.8 Tf [(therefore do not use them.)] TJ ET BT 26.250 360.801 Td /F4 9.8 Tf [(Model)] TJ ET BT 26.250 341.396 Td /F1 9.8 Tf [(When forecasting influenza rates, we used a basic linear autoregressive model for ease of comparison to previous work. )] TJ ET 0.267 0.267 0.267 rg BT 547.007 342.903 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 556.645 345.284 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 559.054 342.903 Td /F4 8.7 Tf [(31)] TJ ET 0.271 0.267 0.267 rg BT 26.250 329.491 Td /F1 9.8 Tf [(Our linear model took the form:)] TJ ET q 212.250 0 0 15.000 26.250 304.610 cm /I16 Do Q BT 26.250 287.586 Td /F1 9.8 Tf [(where y)] TJ ET BT 60.385 285.522 Td /F1 8.7 Tf [(w )] TJ ET BT 69.051 287.586 Td /F1 9.8 Tf [(denotes the ILI prevalence at week w and the values of a are the regression coefficients. When k=0, we are )] TJ ET BT 26.250 275.682 Td /F1 9.8 Tf [(nowcasting inferring the present influenza prevalence rate that the CDC will report in the following week. )] TJ ET 0.267 0.267 0.267 rg BT 492.271 277.189 Td /F4 8.7 Tf [(7)] TJ ET 0.271 0.267 0.267 rg BT 497.089 275.682 Td /F1 9.8 Tf [( When k > 0 we )] TJ ET BT 26.250 263.777 Td /F1 9.8 Tf [(are forecasting further into the future. The )] TJ ET q 8.250 0 0 7.500 208.887 265.801 cm /I18 Do Q BT 217.137 263.777 Td /F1 9.8 Tf [( parameters can be estimated using least-squares regression, where parameters )] TJ ET BT 26.250 251.872 Td /F1 9.8 Tf [(are estimated separately for different values of k.)] TJ ET BT 26.250 232.467 Td /F1 9.8 Tf [(We distinguished between the gold standard data that we are trying to estimate i.e., ILINet data that are no longer being )] TJ ET BT 26.250 220.563 Td /F1 9.8 Tf [(revised and the data that are actually available on week w-1 from the CDCs weekly reports. Whereas y denotes the weekly )] TJ ET BT 26.250 208.582 Td /F1 9.8 Tf [(value in the final gold standard report, )] TJ ET q 10.500 0 0 12.000 198.581 206.182 cm /I20 Do Q BT 209.081 208.582 Td /F1 9.8 Tf [( denotes the values that are published in the report that is most recent at the time of )] TJ ET BT 26.250 196.582 Td /F1 9.8 Tf [(the forecast. Our use of )] TJ ET q 10.500 0 0 12.000 130.848 194.182 cm /I22 Do Q BT 141.348 196.582 Td /F1 9.8 Tf [( in our model ensures that we more accurately reflect the expected performance of forecasts )] TJ ET BT 26.250 184.658 Td /F1 9.8 Tf [(produced using the most recent ILINet data. When forecasting, we trained our baseline model on three seasons of data, )] TJ ET BT 26.250 172.753 Td /F1 9.8 Tf [(beginning with the 2011-2012 season and evaluated it using cross-fold validation, where each season of data is a cross-)] TJ ET BT 26.250 160.848 Td /F1 9.8 Tf [(validation fold. \(We also trained with data beginning in 2004, yielding a negligible reduction in error: a decrease of 0.001.\))] TJ ET BT 26.250 141.444 Td /F1 9.8 Tf [(One advantage of both Twitter and GFT-based systems is that data are available for the current week, not just the previous )] TJ ET BT 26.250 129.539 Td /F1 9.8 Tf [(week. To augment the forecasting model, we include z)] TJ ET BT 261.440 127.475 Td /F1 8.7 Tf [(w)] TJ ET BT 267.697 129.539 Td /F1 9.8 Tf [(, the Web-based methods estimate for week w:)] TJ ET q 250.500 0 0 15.000 26.250 104.658 cm /I24 Do Q BT 26.250 87.634 Td /F1 9.8 Tf [(Similarly, we can utilize multiple Web-based information sources by adding additional terms for each z)] TJ ET BT 465.692 85.570 Td /F1 8.7 Tf [(w)] TJ ET BT 471.950 87.634 Td /F1 9.8 Tf [(.)] TJ ET BT 26.250 67.553 Td /F1 9.8 Tf [(For the purpose of evaluation, we also experimented with a model that uses )] TJ ET BT 355.741 67.553 Td /F5 9.8 Tf [(only)] TJ ET BT 373.623 67.553 Td /F1 9.8 Tf [( Web-based data, )] TJ ET q 54.750 0 0 12.750 453.290 65.003 cm /I26 Do Q BT 508.040 67.553 Td /F1 9.8 Tf [(, which fits the )] TJ ET BT 26.250 55.479 Td /F1 9.8 Tf [(web-based data to the corresponding ILINet values.)] TJ ET Q q 15.000 45.598 577.500 731.402 re W n 0.271 0.267 0.267 rg BT 26.250 750.278 Td /F4 12.0 Tf [(Methods)] TJ ET BT 26.250 730.324 Td /F4 9.8 Tf [(ILINet)] TJ ET BT 26.250 710.919 Td /F1 9.8 Tf [(Our baseline surveillance data and gold standard for predictions are based on CDCs ILINet. Historical and current ILINet data )] TJ ET BT 26.250 699.015 Td /F1 9.8 Tf [(\(at the national, HHS, and Census-division regional levels\) are available from: )] TJ ET 0.267 0.267 0.267 rg BT 26.250 687.110 Td /F1 9.8 Tf [(http://gis.cdc.gov/grasp/fluview/fluportaldashboard.html)] TJ ET 0.271 0.267 0.267 rg BT 263.604 687.110 Td /F1 9.8 Tf [(. Our analyses use the weighted version of this metric, which adjusts for )] TJ ET BT 26.250 675.205 Td /F1 9.8 Tf [(state population. Data are available starting with the 1997-1998 influenza season. Weekly tables released on week )] TJ ET BT 523.207 675.205 Td /F4 9.8 Tf [(W)] TJ ET BT 532.411 675.205 Td /F1 9.8 Tf [( of the )] TJ ET BT 562.227 675.205 Td /F4 9.8 Tf [(X-Y )] TJ ET BT 26.250 663.300 Td /F1 9.8 Tf [(season are available at the following URLs:)] TJ ET BT 26.250 643.896 Td /F1 9.8 Tf [(http://www.cdc.gov/flu/weekly/weeklyarchives)] TJ ET BT 222.391 643.896 Td /F4 9.8 Tf [(X-Y)] TJ ET BT 238.644 643.896 Td /F1 9.8 Tf [(/data/senAllregt)] TJ ET BT 306.387 643.896 Td /F4 9.8 Tf [(W)] TJ ET BT 315.591 643.896 Td /F1 9.8 Tf [(.htm.)] TJ ET BT 26.250 624.491 Td /F1 9.8 Tf [(In addition to final ILINet values, we downloaded ILINet data that were available at a particular time from these tables. Such )] TJ ET BT 26.250 612.586 Td /F1 9.8 Tf [(tables are available for all seasons beginning 2004-2005.)] TJ ET BT 26.250 593.181 Td /F4 9.8 Tf [(Twitter)] TJ ET BT 26.250 573.777 Td /F1 9.8 Tf [(We use the Twitter influenza surveillance system developed by Lamb et al. )] TJ ET 0.267 0.267 0.267 rg BT 351.383 575.284 Td /F4 8.7 Tf [(11)] TJ ET 0.271 0.267 0.267 rg BT 361.021 577.665 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 363.430 575.284 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 373.067 573.777 Td /F1 9.8 Tf [( to produce weekly influenza rates, since it )] TJ ET BT 26.250 561.872 Td /F1 9.8 Tf [(achieves state of the art surveillance results for Twitter data. This algorithm identifies such messages using a cascade of logistic )] TJ ET BT 26.250 549.967 Td /F1 9.8 Tf [(regression classifiers that determine, first, if a message is about health; next, if it is about influenza; and finally, if it is about an )] TJ ET BT 26.250 538.062 Td /F1 9.8 Tf [(influenza infection \(rather than simply an awareness of the ongoing flu season\). The classifiers are trained purely on the )] TJ ET BT 26.250 526.158 Td /F1 9.8 Tf [(message content rather than on historical ILI data. We used messages from the United States as determined by the )] TJ ET BT 527.536 526.158 Td /F5 9.8 Tf [(Carmen)] TJ ET BT 26.250 514.253 Td /F1 9.8 Tf [(geolocation system. )] TJ ET 0.267 0.267 0.267 rg BT 115.121 515.760 Td /F4 8.7 Tf [(27)] TJ ET 0.271 0.267 0.267 rg BT 124.759 514.253 Td /F1 9.8 Tf [( We use the output of these models as features for the forecast model based on data from November 27, )] TJ ET BT 26.250 502.348 Td /F1 9.8 Tf [(2011 through April 5, 2014.)] TJ ET BT 26.250 482.943 Td /F4 9.8 Tf [(Google Flu Trends)] TJ ET BT 26.250 463.539 Td /F1 9.8 Tf [(Google Flu Trends is an influenza surveillance system that estimates current infection rates based on the volume of Google )] TJ ET BT 26.250 451.634 Td /F1 9.8 Tf [(searches for a select number of influenza related queries. )] TJ ET 0.267 0.267 0.267 rg BT 277.156 453.141 Td /F4 8.7 Tf [(13)] TJ ET 0.271 0.267 0.267 rg BT 286.794 451.634 Td /F1 9.8 Tf [( GFT data are available from )] TJ ET 0.267 0.267 0.267 rg BT 26.250 439.729 Td /F1 9.8 Tf [(http://www.google.org/flutrends/us/data.txt)] TJ ET 0.271 0.267 0.267 rg BT 208.341 439.729 Td /F1 9.8 Tf [(. We collected all GFT estimates for the US and restrict our attention to the same time )] TJ ET BT 26.250 427.824 Td /F1 9.8 Tf [(interval as our Twitter data for a direct comparison. Following poor performance during the 2012-13 season, )] TJ ET 0.267 0.267 0.267 rg BT 493.363 429.332 Td /F4 8.7 Tf [(21)] TJ ET 0.271 0.267 0.267 rg BT 503.000 427.824 Td /F1 9.8 Tf [( the GFT model )] TJ ET BT 26.250 415.920 Td /F1 9.8 Tf [(was updated in August 2013. Therefore, the numbers beginning in this month are based on the latest model. Retrospective )] TJ ET BT 26.250 404.015 Td /F1 9.8 Tf [(estimates are also available for earlier data using the newest model, but because these estimates are based on a model trained )] TJ ET BT 26.250 392.110 Td /F1 9.8 Tf [(on the same data \(in-sample data\), they do not provide an accurate assessment of the models predictive abilities and we )] TJ ET BT 26.250 380.205 Td /F1 9.8 Tf [(therefore do not use them.)] TJ ET BT 26.250 360.801 Td /F4 9.8 Tf [(Model)] TJ ET BT 26.250 341.396 Td /F1 9.8 Tf [(When forecasting influenza rates, we used a basic linear autoregressive model for ease of comparison to previous work. )] TJ ET 0.267 0.267 0.267 rg BT 547.007 342.903 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 556.645 345.284 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 559.054 342.903 Td /F4 8.7 Tf [(31)] TJ ET 0.271 0.267 0.267 rg BT 26.250 329.491 Td /F1 9.8 Tf [(Our linear model took the form:)] TJ ET q 212.250 0 0 15.000 26.250 304.610 cm /I28 Do Q BT 26.250 287.586 Td /F1 9.8 Tf [(where y)] TJ ET BT 60.385 285.522 Td /F1 8.7 Tf [(w )] TJ ET BT 69.051 287.586 Td /F1 9.8 Tf [(denotes the ILI prevalence at week w and the values of a are the regression coefficients. When k=0, we are )] TJ ET BT 26.250 275.682 Td /F1 9.8 Tf [(nowcasting inferring the present influenza prevalence rate that the CDC will report in the following week. )] TJ ET 0.267 0.267 0.267 rg BT 492.271 277.189 Td /F4 8.7 Tf [(7)] TJ ET 0.271 0.267 0.267 rg BT 497.089 275.682 Td /F1 9.8 Tf [( When k > 0 we )] TJ ET BT 26.250 263.777 Td /F1 9.8 Tf [(are forecasting further into the future. The )] TJ ET q 8.250 0 0 7.500 208.887 265.801 cm /I30 Do Q BT 217.137 263.777 Td /F1 9.8 Tf [( parameters can be estimated using least-squares regression, where parameters )] TJ ET BT 26.250 251.872 Td /F1 9.8 Tf [(are estimated separately for different values of k.)] TJ ET BT 26.250 232.467 Td /F1 9.8 Tf [(We distinguished between the gold standard data that we are trying to estimate i.e., ILINet data that are no longer being )] TJ ET BT 26.250 220.563 Td /F1 9.8 Tf [(revised and the data that are actually available on week w-1 from the CDCs weekly reports. Whereas y denotes the weekly )] TJ ET BT 26.250 208.582 Td /F1 9.8 Tf [(value in the final gold standard report, )] TJ ET q 10.500 0 0 12.000 198.581 206.182 cm /I32 Do Q BT 209.081 208.582 Td /F1 9.8 Tf [( denotes the values that are published in the report that is most recent at the time of )] TJ ET BT 26.250 196.582 Td /F1 9.8 Tf [(the forecast. Our use of )] TJ ET q 10.500 0 0 12.000 130.848 194.182 cm /I34 Do Q BT 141.348 196.582 Td /F1 9.8 Tf [( in our model ensures that we more accurately reflect the expected performance of forecasts )] TJ ET BT 26.250 184.658 Td /F1 9.8 Tf [(produced using the most recent ILINet data. When forecasting, we trained our baseline model on three seasons of data, )] TJ ET BT 26.250 172.753 Td /F1 9.8 Tf [(beginning with the 2011-2012 season and evaluated it using cross-fold validation, where each season of data is a cross-)] TJ ET BT 26.250 160.848 Td /F1 9.8 Tf [(validation fold. \(We also trained with data beginning in 2004, yielding a negligible reduction in error: a decrease of 0.001.\))] TJ ET BT 26.250 141.444 Td /F1 9.8 Tf [(One advantage of both Twitter and GFT-based systems is that data are available for the current week, not just the previous )] TJ ET BT 26.250 129.539 Td /F1 9.8 Tf [(week. To augment the forecasting model, we include z)] TJ ET BT 261.440 127.475 Td /F1 8.7 Tf [(w)] TJ ET BT 267.697 129.539 Td /F1 9.8 Tf [(, the Web-based methods estimate for week w:)] TJ ET q 250.500 0 0 15.000 26.250 104.658 cm /I36 Do Q BT 26.250 87.634 Td /F1 9.8 Tf [(Similarly, we can utilize multiple Web-based information sources by adding additional terms for each z)] TJ ET BT 465.692 85.570 Td /F1 8.7 Tf [(w)] TJ ET BT 471.950 87.634 Td /F1 9.8 Tf [(.)] TJ ET BT 26.250 67.553 Td /F1 9.8 Tf [(For the purpose of evaluation, we also experimented with a model that uses )] TJ ET BT 355.741 67.553 Td /F5 9.8 Tf [(only)] TJ ET BT 373.623 67.553 Td /F1 9.8 Tf [( Web-based data, )] TJ ET q 54.750 0 0 12.750 453.290 65.003 cm /I38 Do Q BT 508.040 67.553 Td /F1 9.8 Tf [(, which fits the )] TJ ET BT 26.250 55.479 Td /F1 9.8 Tf [(web-based data to the corresponding ILINet values.)] TJ ET Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(2)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj 238 0 obj << /Type /Font /Subtype /Type1 /Name /F5 /BaseFont /Helvetica-Oblique /Encoding /WinAnsiEncoding >> endobj 239 0 obj << /Type /Annot /Subtype /Link /A 240 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 686.2081 263.6040 696.1287 ] >> endobj 240 0 obj << /Type /Action /S /URI /URI (http://gis.cdc.gov/grasp/fluview/fluportaldashboard.html) >> endobj 241 0 obj << /Type /Annot /Subtype /Link /A 242 0 R /Border [0 0 0] /H /I /Rect [ 351.3832 574.4822 361.0206 583.3005 ] >> endobj 242 0 obj << /Type /Action >> endobj 243 0 obj << /Type /Annot /Subtype /Link /A 244 0 R /Border [0 0 0] /H /I /Rect [ 363.4299 574.4822 373.0673 583.3005 ] >> endobj 244 0 obj << /Type /Action >> endobj 245 0 obj << /Type /Annot /Subtype /Link /A 246 0 R /Border [0 0 0] /H /I /Rect [ 115.1213 514.9584 124.7586 523.7767 ] >> endobj 246 0 obj << /Type /Action >> endobj 247 0 obj << /Type /Annot /Subtype /Link /A 248 0 R /Border [0 0 0] /H /I /Rect [ 277.1565 452.3394 286.7938 461.1578 ] >> endobj 248 0 obj << /Type /Action >> endobj 249 0 obj << /Type /Annot /Subtype /Link /A 250 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 438.8273 208.3410 448.7480 ] >> endobj 250 0 obj << /Type /Action /S /URI /URI (http://www.google.org/flutrends/us/data.txt) >> endobj 251 0 obj << /Type /Annot /Subtype /Link /A 252 0 R /Border [0 0 0] /H /I /Rect [ 493.3628 428.5299 503.0001 437.3483 ] >> endobj 252 0 obj << /Type /Action >> endobj 253 0 obj << /Type /Annot /Subtype /Link /A 254 0 R /Border [0 0 0] /H /I /Rect [ 547.0072 342.1014 556.6446 350.9198 ] >> endobj 254 0 obj << /Type /Action >> endobj 255 0 obj << /Type /Annot /Subtype /Link /A 256 0 R /Border [0 0 0] /H /I /Rect [ 559.0539 342.1014 568.6913 350.9198 ] >> endobj 256 0 obj << /Type /Action >> endobj 257 0 obj << /Type /XObject /Subtype /Image /Width 283 /Height 20 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 283 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1281>> stream XXg+[^ڮ-*#*t*uE81ꨈ>\ǨqcQTcqUDT䡪*]{gϙL̷k=1ʯ뢣1q#[?ahG-F5]Mό;m_\p} 0Bwsi .XX>XZشAmhl&Ɣ63E 7?/B<mw( !_jW2}9-a8:afֽ(zEg^V٣" A^w+(G^BtñtHdʴ`הavE#\Ok]Je$8OϓhQ,Qq|C_kʰr7CN[S(ֺO8ETtў TIߍș|m,"}?J hsIJj<|9{8g1'+ I#o \Ty*dA2NobUO>m<|mpaQ.6*ے01"6ƧM &Fʸ[,YkW#`_l^Ys(pLyH"m`fG7JnPDJh EIBm|pKg EuGΗ&W9=n#bm\pKw}U[zر&PP" C?y,Z/@N*jaCc;!]E' YB ~\ݻ+(5.)`.?I˸8ZMcMB(ό!cM:M86[;X=iI\׫4Ev Y_ jissY$/Xw(i>8^cH(Bv0!U P|6jOVچʭy[nDn(B]q9ӌ:0 [$g-}VyEP ڰ ΡZqɴ u:Q ]cl=棣i@&o5Pwx2I Ʃ ROAk% ~+>@-lUJAu kul=L?iJB. endstream endobj 258 0 obj << /Type /XObject /Subtype /Image /Width 283 /Height 20 /SMask 257 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 283 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 38>> stream x nH@`Bh endstream endobj 259 0 obj << /Type /Annot /Subtype /Link /A 260 0 R /Border [0 0 0] /H /I /Rect [ 492.2708 276.3872 497.0894 285.2055 ] >> endobj 260 0 obj << /Type /Action >> endobj 261 0 obj << /Type /XObject /Subtype /Image /Width 11 /Height 10 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 11 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 106>> stream c`&6(X7 LX'/a )Uz kԵhPb:1 >Rq "gLA+v010,5)LfϤ `b[ endstream endobj 262 0 obj << /Type /XObject /Subtype /Image /Width 11 /Height 10 /SMask 261 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 11 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 13>> stream c`T endstream endobj 263 0 obj << /Type /XObject /Subtype /Image /Width 14 /Height 16 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 14 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 148>> stream c`6f3MP! R c``˙W&> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 15>> stream (c`` endstream endobj 265 0 obj << /Type /XObject /Subtype /Image /Width 14 /Height 16 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 14 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 148>> stream c`6f3MP! R c``˙W&> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 15>> stream (c`` endstream endobj 267 0 obj << /Type /XObject /Subtype /Image /Width 334 /Height 20 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 334 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1912>> stream XXg+K?ZbEETDUDUUEEUDE=WTDUQUQZQUQ+"VDTTDTEEDDTTUTT'mvm7޻fvfΜ6 _!LMhwJCLW?h ]n&3YLɆ˵Nk돻Vz -(_Fv_K8V @sbAqUa%`9(ۂk"f@G0b}kYd.ږFM2L8d l ڹ9]EU!bj$]'C j~HRhH!1*}[>YO$Ro2\3|UUZFýkzU[ػm\HYj7ޯD_T8hxMn2%w2AE(eQll+ҮؗMK?*)9pJ:]r(p#ک6pJ* ?]'}W H8cm8ܖt-9_ i%3\gUɲ(#yr mQw,dA-+mb-+:M*ut 9y쨭(!a|NN)7er{Li"{GjR} W2 ût:Cqv .1|x%q nx^җOzL$*DUu5pj"a&=䚘B* !D>Qtfa;n![Գ|:#6.[& ?%o8KP-9_\#8I*.^'UyӅ GRNiq-G[\'%+r^IrdԞ^qO 杄 E>=)oʳܵwuS/WwN߰ >֣@maI'yTػ/#I+pPظmk1@rS6a vGXoMڣ5IF䇈ЍC^:Yz~0\;RU/5|||FW1}vsը6PCQr,\4oC~7lɄݷV;l\k)Yq BdIiI36?"XELS$Sr؟9݇9I`w, %:e'>Bxj(Ԇ/0k&wеh)> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 42>> stream x  Om7x7N\ endstream endobj 269 0 obj << /Type /XObject /Subtype /Image /Width 73 /Height 17 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 73 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 443>> stream 8T1K@}Jg$89"YJ(8"EDB"C "BDDJp)E8H)EJHմ}]{w}G0BQ*.w49T)~*?ԋwޜ>e ޛ.|?EU:3w>IӐo%$n3I*I ?3F[r]&IP7Ysc~M>L혈Jm"+7ޥ O)mezyzĠv=E VܦAv6~Ӏ5ֵ`V!G 9x mrÄnءhWh/MmyL93}'i&)%^ijȝ1^ -z-ψpK'3]nVAN `~0dT_xA%,V}@*.жxU endstream endobj 270 0 obj << /Type /XObject /Subtype /Image /Width 73 /Height 17 /SMask 269 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 73 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 26>> stream H1 Om k endstream endobj 271 0 obj << /Type /Annot /Subtype /Link /A 272 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 686.2081 263.6040 696.1287 ] >> endobj 272 0 obj << /Type /Action /S /URI /URI (http://gis.cdc.gov/grasp/fluview/fluportaldashboard.html) >> endobj 273 0 obj << /Type /Annot /Subtype /Link /A 274 0 R /Border [0 0 0] /H /I /Rect [ 351.3832 574.4822 361.0206 583.3005 ] >> endobj 274 0 obj << /Type /Action >> endobj 275 0 obj << /Type /Annot /Subtype /Link /A 276 0 R /Border [0 0 0] /H /I /Rect [ 363.4299 574.4822 373.0673 583.3005 ] >> endobj 276 0 obj << /Type /Action >> endobj 277 0 obj << /Type /Annot /Subtype /Link /A 278 0 R /Border [0 0 0] /H /I /Rect [ 115.1213 514.9584 124.7586 523.7767 ] >> endobj 278 0 obj << /Type /Action >> endobj 279 0 obj << /Type /Annot /Subtype /Link /A 280 0 R /Border [0 0 0] /H /I /Rect [ 277.1565 452.3394 286.7938 461.1578 ] >> endobj 280 0 obj << /Type /Action >> endobj 281 0 obj << /Type /Annot /Subtype /Link /A 282 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 438.8273 208.3410 448.7480 ] >> endobj 282 0 obj << /Type /Action /S /URI /URI (http://www.google.org/flutrends/us/data.txt) >> endobj 283 0 obj << /Type /Annot /Subtype /Link /A 284 0 R /Border [0 0 0] /H /I /Rect [ 493.3628 428.5299 503.0001 437.3483 ] >> endobj 284 0 obj << /Type /Action >> endobj 285 0 obj << /Type /Annot /Subtype /Link /A 286 0 R /Border [0 0 0] /H /I /Rect [ 547.0072 342.1014 556.6446 350.9198 ] >> endobj 286 0 obj << /Type /Action >> endobj 287 0 obj << /Type /Annot /Subtype /Link /A 288 0 R /Border [0 0 0] /H /I /Rect [ 559.0539 342.1014 568.6913 350.9198 ] >> endobj 288 0 obj << /Type /Action >> endobj 289 0 obj << /Type /XObject /Subtype /Image /Width 283 /Height 20 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 283 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1281>> stream XXg+[^ڮ-*#*t*uE81ꨈ>\ǨqcQTcqUDT䡪*]{gϙL̷k=1ʯ뢣1q#[?ahG-F5]Mό;m_\p} 0Bwsi .XX>XZشAmhl&Ɣ63E 7?/B<mw( !_jW2}9-a8:afֽ(zEg^V٣" A^w+(G^BtñtHdʴ`הavE#\Ok]Je$8OϓhQ,Qq|C_kʰr7CN[S(ֺO8ETtў TIߍș|m,"}?J hsIJj<|9{8g1'+ I#o \Ty*dA2NobUO>m<|mpaQ.6*ے01"6ƧM &Fʸ[,YkW#`_l^Ys(pLyH"m`fG7JnPDJh EIBm|pKg EuGΗ&W9=n#bm\pKw}U[zر&PP" C?y,Z/@N*jaCc;!]E' YB ~\ݻ+(5.)`.?I˸8ZMcMB(ό!cM:M86[;X=iI\׫4Ev Y_ jissY$/Xw(i>8^cH(Bv0!U P|6jOVچʭy[nDn(B]q9ӌ:0 [$g-}VyEP ڰ ΡZqɴ u:Q ]cl=棣i@&o5Pwx2I Ʃ ROAk% ~+>@-lUJAu kul=L?iJB. endstream endobj 290 0 obj << /Type /XObject /Subtype /Image /Width 283 /Height 20 /SMask 289 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 283 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 38>> stream x nH@`Bh endstream endobj 291 0 obj << /Type /Annot /Subtype /Link /A 292 0 R /Border [0 0 0] /H /I /Rect [ 492.2708 276.3872 497.0894 285.2055 ] >> endobj 292 0 obj << /Type /Action >> endobj 293 0 obj << /Type /XObject /Subtype /Image /Width 11 /Height 10 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 11 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 106>> stream c`&6(X7 LX'/a )Uz kԵhPb:1 >Rq "gLA+v010,5)LfϤ `b[ endstream endobj 294 0 obj << /Type /XObject /Subtype /Image /Width 11 /Height 10 /SMask 293 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 11 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 13>> stream c`T endstream endobj 295 0 obj << /Type /XObject /Subtype /Image /Width 14 /Height 16 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 14 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 148>> stream c`6f3MP! R c``˙W&> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 15>> stream (c`` endstream endobj 297 0 obj << /Type /XObject /Subtype /Image /Width 14 /Height 16 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 14 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 148>> stream c`6f3MP! R c``˙W&> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 15>> stream (c`` endstream endobj 299 0 obj << /Type /XObject /Subtype /Image /Width 334 /Height 20 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 334 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1912>> stream XXg+K?ZbEETDUDUUEEUDE=WTDUQUQZQUQ+"VDTTDTEEDDTTUTT'mvm7޻fvfΜ6 _!LMhwJCLW?h ]n&3YLɆ˵Nk돻Vz -(_Fv_K8V @sbAqUa%`9(ۂk"f@G0b}kYd.ږFM2L8d l ڹ9]EU!bj$]'C j~HRhH!1*}[>YO$Ro2\3|UUZFýkzU[ػm\HYj7ޯD_T8hxMn2%w2AE(eQll+ҮؗMK?*)9pJ:]r(p#ک6pJ* ?]'}W H8cm8ܖt-9_ i%3\gUɲ(#yr mQw,dA-+mb-+:M*ut 9y쨭(!a|NN)7er{Li"{GjR} W2 ût:Cqv .1|x%q nx^җOzL$*DUu5pj"a&=䚘B* !D>Qtfa;n![Գ|:#6.[& ?%o8KP-9_\#8I*.^'UyӅ GRNiq-G[\'%+r^IrdԞ^qO 杄 E>=)oʳܵwuS/WwN߰ >֣@maI'yTػ/#I+pPظmk1@rS6a vGXoMڣ5IF䇈ЍC^:Yz~0\;RU/5|||FW1}vsը6PCQr,\4oC~7lɄݷV;l\k)Yq BdIiI36?"XELS$Sr؟9݇9I`w, %:e'>Bxj(Ԇ/0k&wеh)> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 42>> stream x  Om7x7N\ endstream endobj 301 0 obj << /Type /XObject /Subtype /Image /Width 73 /Height 17 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 73 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 443>> stream 8T1K@}Jg$89"YJ(8"EDB"C "BDDJp)E8H)EJHմ}]{w}G0BQ*.w49T)~*?ԋwޜ>e ޛ.|?EU:3w>IӐo%$n3I*I ?3F[r]&IP7Ysc~M>L혈Jm"+7ޥ O)mezyzĠv=E VܦAv6~Ӏ5ֵ`V!G 9x mrÄnءhWh/MmyL93}'i&)%^ijȝ1^ -z-ψpK'3]nVAN `~0dT_xA%,V}@*.жxU endstream endobj 302 0 obj << /Type /XObject /Subtype /Image /Width 73 /Height 17 /SMask 301 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 73 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 26>> stream H1 Om k endstream endobj 303 0 obj << /Type /Annot /Subtype /Link /A 304 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 686.2081 263.6040 696.1287 ] >> endobj 304 0 obj << /Type /Action /S /URI /URI (http://gis.cdc.gov/grasp/fluview/fluportaldashboard.html) >> endobj 305 0 obj << /Type /Annot /Subtype /Link /A 306 0 R /Border [0 0 0] /H /I /Rect [ 351.3832 574.4822 361.0206 583.3005 ] >> endobj 306 0 obj << /Type /Action >> endobj 307 0 obj << /Type /Annot /Subtype /Link /A 308 0 R /Border [0 0 0] /H /I /Rect [ 363.4299 574.4822 373.0673 583.3005 ] >> endobj 308 0 obj << /Type /Action >> endobj 309 0 obj << /Type /Annot /Subtype /Link /A 310 0 R /Border [0 0 0] /H /I /Rect [ 115.1213 514.9584 124.7586 523.7767 ] >> endobj 310 0 obj << /Type /Action >> endobj 311 0 obj << /Type /Annot /Subtype /Link /A 312 0 R /Border [0 0 0] /H /I /Rect [ 277.1565 452.3394 286.7938 461.1578 ] >> endobj 312 0 obj << /Type /Action >> endobj 313 0 obj << /Type /Annot /Subtype /Link /A 314 0 R /Border [0 0 0] /H /I /Rect [ 26.2500 438.8273 208.3410 448.7480 ] >> endobj 314 0 obj << /Type /Action /S /URI /URI (http://www.google.org/flutrends/us/data.txt) >> endobj 315 0 obj << /Type /Annot /Subtype /Link /A 316 0 R /Border [0 0 0] /H /I /Rect [ 493.3628 428.5299 503.0001 437.3483 ] >> endobj 316 0 obj << /Type /Action >> endobj 317 0 obj << /Type /Annot /Subtype /Link /A 318 0 R /Border [0 0 0] /H /I /Rect [ 547.0072 342.1014 556.6446 350.9198 ] >> endobj 318 0 obj << /Type /Action >> endobj 319 0 obj << /Type /Annot /Subtype /Link /A 320 0 R /Border [0 0 0] /H /I /Rect [ 559.0539 342.1014 568.6913 350.9198 ] >> endobj 320 0 obj << /Type /Action >> endobj 321 0 obj << /Type /XObject /Subtype /Image /Width 283 /Height 20 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 283 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1281>> stream XXg+[^ڮ-*#*t*uE81ꨈ>\ǨqcQTcqUDT䡪*]{gϙL̷k=1ʯ뢣1q#[?ahG-F5]Mό;m_\p} 0Bwsi .XX>XZشAmhl&Ɣ63E 7?/B<mw( !_jW2}9-a8:afֽ(zEg^V٣" A^w+(G^BtñtHdʴ`הavE#\Ok]Je$8OϓhQ,Qq|C_kʰr7CN[S(ֺO8ETtў TIߍș|m,"}?J hsIJj<|9{8g1'+ I#o \Ty*dA2NobUO>m<|mpaQ.6*ے01"6ƧM &Fʸ[,YkW#`_l^Ys(pLyH"m`fG7JnPDJh EIBm|pKg EuGΗ&W9=n#bm\pKw}U[zر&PP" C?y,Z/@N*jaCc;!]E' YB ~\ݻ+(5.)`.?I˸8ZMcMB(ό!cM:M86[;X=iI\׫4Ev Y_ jissY$/Xw(i>8^cH(Bv0!U P|6jOVچʭy[nDn(B]q9ӌ:0 [$g-}VyEP ڰ ΡZqɴ u:Q ]cl=棣i@&o5Pwx2I Ʃ ROAk% ~+>@-lUJAu kul=L?iJB. endstream endobj 322 0 obj << /Type /XObject /Subtype /Image /Width 283 /Height 20 /SMask 321 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 283 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 38>> stream x nH@`Bh endstream endobj 323 0 obj << /Type /Annot /Subtype /Link /A 324 0 R /Border [0 0 0] /H /I /Rect [ 492.2708 276.3872 497.0894 285.2055 ] >> endobj 324 0 obj << /Type /Action >> endobj 325 0 obj << /Type /XObject /Subtype /Image /Width 11 /Height 10 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 11 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 106>> stream c`&6(X7 LX'/a )Uz kԵhPb:1 >Rq "gLA+v010,5)LfϤ `b[ endstream endobj 326 0 obj << /Type /XObject /Subtype /Image /Width 11 /Height 10 /SMask 325 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 11 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 13>> stream c`T endstream endobj 327 0 obj << /Type /XObject /Subtype /Image /Width 14 /Height 16 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 14 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 148>> stream c`6f3MP! R c``˙W&> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 15>> stream (c`` endstream endobj 329 0 obj << /Type /XObject /Subtype /Image /Width 14 /Height 16 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 14 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 148>> stream c`6f3MP! R c``˙W&> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 15>> stream (c`` endstream endobj 331 0 obj << /Type /XObject /Subtype /Image /Width 334 /Height 20 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 334 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1912>> stream XXg+K?ZbEETDUDUUEEUDE=WTDUQUQZQUQ+"VDTTDTEEDDTTUTT'mvm7޻fvfΜ6 _!LMhwJCLW?h ]n&3YLɆ˵Nk돻Vz -(_Fv_K8V @sbAqUa%`9(ۂk"f@G0b}kYd.ږFM2L8d l ڹ9]EU!bj$]'C j~HRhH!1*}[>YO$Ro2\3|UUZFýkzU[ػm\HYj7ޯD_T8hxMn2%w2AE(eQll+ҮؗMK?*)9pJ:]r(p#ک6pJ* ?]'}W H8cm8ܖt-9_ i%3\gUɲ(#yr mQw,dA-+mb-+:M*ut 9y쨭(!a|NN)7er{Li"{GjR} W2 ût:Cqv .1|x%q nx^җOzL$*DUu5pj"a&=䚘B* !D>Qtfa;n![Գ|:#6.[& ?%o8KP-9_\#8I*.^'UyӅ GRNiq-G[\'%+r^IrdԞ^qO 杄 E>=)oʳܵwuS/WwN߰ >֣@maI'yTػ/#I+pPظmk1@rS6a vGXoMڣ5IF䇈ЍC^:Yz~0\;RU/5|||FW1}vsը6PCQr,\4oC~7lɄݷV;l\k)Yq BdIiI36?"XELS$Sr؟9݇9I`w, %:e'>Bxj(Ԇ/0k&wеh)> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 42>> stream x  Om7x7N\ endstream endobj 333 0 obj << /Type /XObject /Subtype /Image /Width 73 /Height 17 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 73 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 443>> stream 8T1K@}Jg$89"YJ(8"EDB"C "BDDJp)E8H)EJHմ}]{w}G0BQ*.w49T)~*?ԋwޜ>e ޛ.|?EU:3w>IӐo%$n3I*I ?3F[r]&IP7Ysc~M>L혈Jm"+7ޥ O)mezyzĠv=E VܦAv6~Ӏ5ֵ`V!G 9x mrÄnءhWh/MmyL93}'i&)%^ijȝ1^ -z-ψpK'3]nVAN `~0dT_xA%,V}@*.жxU endstream endobj 334 0 obj << /Type /XObject /Subtype /Image /Width 73 /Height 17 /SMask 333 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 73 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 26>> stream H1 Om k endstream endobj 335 0 obj << /Type /Page /Parent 3 0 R /Annots [ 339 0 R 341 0 R 345 0 R 347 0 R 351 0 R 353 0 R ] /Contents 336 0 R >> endobj 336 0 obj << /Length 37444 >> stream 0.271 0.267 0.267 rg q 15.000 168.651 577.500 608.349 re W n 0.271 0.267 0.267 rg BT 26.250 767.476 Td /F1 9.8 Tf [(We include a nonparametric baseline, which predicts each weeks value as the average value from all historical data for that )] TJ ET BT 26.250 755.571 Td /F1 9.8 Tf [(week from 1997-2010. That is,)] TJ ET q 229.500 0 0 24.750 26.250 720.940 cm /I40 Do Q BT 26.250 703.917 Td /F1 9.8 Tf [(where y)] TJ ET BT 60.385 707.805 Td /F1 8.7 Tf [(i)] TJ ET BT 62.309 701.852 Td /F1 8.7 Tf [(w )] TJ ET BT 70.975 703.917 Td /F1 9.8 Tf [(is the value at week w in the season starting in year i and ending in year i+1.)] TJ ET BT 26.250 684.512 Td /F1 9.8 Tf [(The purpose of this comparison is to understand how much information we are gaining using autoregressive models over simply )] TJ ET BT 26.250 672.607 Td /F1 9.8 Tf [(modeling each season as the average of previous seasons.)] TJ ET BT 26.250 636.005 Td /F4 12.0 Tf [(Results)] TJ ET q 26.250 611.838 555.000 13.736 re W n 0.271 0.267 0.267 rg BT 26.250 614.585 Td /F1 9.8 Tf [(Table 1. Mean absolute errors from cross-validation across three seasons for the nowcasting task)] TJ ET Q 0.965 0.965 0.965 rg 26.250 562.615 555.000 41.722 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 604.338 m 581.250 604.338 l 580.500 603.588 l 27.000 603.588 l f 581.250 604.338 m 581.250 562.615 l 580.500 562.615 l 580.500 603.588 l f 26.250 604.338 m 26.250 562.615 l 27.000 562.615 l 27.000 603.588 l f 0.271 0.267 0.267 rg BT 33.000 590.262 Td /F1 9.0 Tf [(\(a\) Final revised CDC weekly estimates; \(b\) the realistic model using original CDC data before revision; \(c\) the model augmented with )] TJ ET BT 33.000 581.104 Td /F1 9.0 Tf [(Twitter data; \(d\) the model augmented with GFT data; \(e\) the model augmented with both Twitter and GFT data; \(f\) values predicted by )] TJ ET BT 33.000 571.947 Td /F1 9.0 Tf [(measuring the historical average.)] TJ ET 1.000 1.000 1.000 rg 26.250 405.904 555.000 156.711 re f 0.965 0.965 0.965 rg 27.000 549.586 50.689 12.280 re f 0.267 0.267 0.267 rg 26.625 561.490 51.064 0.750 re f 26.625 549.211 0.750 13.030 re f 0.965 0.965 0.965 rg 77.689 549.586 280.266 12.280 re f 0.267 0.267 0.267 rg 77.689 561.490 280.266 0.750 re f 0.271 0.267 0.267 rg BT 82.189 552.342 Td /F4 9.8 Tf [(Model)] TJ ET 0.965 0.965 0.965 rg 357.955 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 357.955 561.490 74.182 0.750 re f 0.271 0.267 0.267 rg BT 362.455 552.342 Td /F4 9.8 Tf [(11-12)] TJ ET 0.965 0.965 0.965 rg 432.137 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 432.137 561.490 74.182 0.750 re f 0.271 0.267 0.267 rg BT 436.637 552.342 Td /F4 9.8 Tf [(12-13)] TJ ET 0.965 0.965 0.965 rg 506.318 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 506.318 561.490 74.557 0.750 re f 580.125 549.211 0.750 13.030 re f 0.271 0.267 0.267 rg BT 510.818 552.342 Td /F4 9.8 Tf [(13-14)] TJ ET 0.267 0.267 0.267 rg 26.625 549.211 51.439 0.750 re f 26.625 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 539.356 Td /F1 9.8 Tf [(\(a\))] TJ ET 0.267 0.267 0.267 rg 77.314 549.211 281.016 0.750 re f 77.314 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 539.356 Td /F1 9.8 Tf [(Revised CDC \(y\))] TJ ET 0.267 0.267 0.267 rg 357.580 549.211 74.932 0.750 re f 357.580 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 539.356 Td /F1 9.8 Tf [(0.10)] TJ ET 0.267 0.267 0.267 rg 431.762 549.211 74.932 0.750 re f 431.762 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 539.356 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 505.943 549.211 74.932 0.750 re f 505.943 533.329 0.750 16.631 re f 580.125 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 539.356 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 26.625 533.329 51.439 0.750 re f 26.625 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 523.474 Td /F1 9.8 Tf [(\(b\))] TJ ET 0.267 0.267 0.267 rg 77.314 533.329 281.016 0.750 re f 77.314 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 523.474 Td /F1 9.8 Tf [(Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 533.329 74.932 0.750 re f 357.580 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 523.474 Td /F1 9.8 Tf [(0.20)] TJ ET 0.267 0.267 0.267 rg 431.762 533.329 74.932 0.750 re f 431.762 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 523.474 Td /F1 9.8 Tf [(0.30)] TJ ET 0.267 0.267 0.267 rg 505.943 533.329 74.932 0.750 re f 505.943 517.448 0.750 16.631 re f 580.125 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 523.474 Td /F1 9.8 Tf [(0.32)] TJ ET 0.267 0.267 0.267 rg 26.625 517.448 51.439 0.750 re f 26.625 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 507.593 Td /F1 9.8 Tf [(\(c\))] TJ ET 0.267 0.267 0.267 rg 77.314 517.448 281.016 0.750 re f 77.314 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 507.593 Td /F1 9.8 Tf [(Twitter \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 517.448 74.932 0.750 re f 357.580 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 507.593 Td /F1 9.8 Tf [(0.33)] TJ ET 0.267 0.267 0.267 rg 431.762 517.448 74.932 0.750 re f 431.762 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 507.593 Td /F1 9.8 Tf [(0.36)] TJ ET 0.267 0.267 0.267 rg 505.943 517.448 74.932 0.750 re f 505.943 501.567 0.750 16.631 re f 580.125 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 507.593 Td /F1 9.8 Tf [(0.48)] TJ ET 0.267 0.267 0.267 rg 26.625 501.567 51.439 0.750 re f 26.625 485.686 0.750 16.631 re f 77.314 501.567 281.016 0.750 re f 77.314 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 491.712 Td /F1 9.8 Tf [(Twitter \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 501.567 74.932 0.750 re f 357.580 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 491.712 Td /F1 9.8 Tf [(0.14)] TJ ET 0.267 0.267 0.267 rg 431.762 501.567 74.932 0.750 re f 431.762 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 491.712 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 505.943 501.567 74.932 0.750 re f 505.943 485.686 0.750 16.631 re f 580.125 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 491.712 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 26.625 485.686 51.439 0.750 re f 26.625 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 475.831 Td /F1 9.8 Tf [(\(d\))] TJ ET 0.267 0.267 0.267 rg 77.314 485.686 281.016 0.750 re f 77.314 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 475.831 Td /F1 9.8 Tf [(GFT \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 485.686 74.932 0.750 re f 357.580 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 475.831 Td /F1 9.8 Tf [(0.35)] TJ ET 0.267 0.267 0.267 rg 431.762 485.686 74.932 0.750 re f 431.762 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 475.831 Td /F1 9.8 Tf [(0.71)] TJ ET 0.267 0.267 0.267 rg 505.943 485.686 74.932 0.750 re f 505.943 469.804 0.750 16.631 re f 580.125 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 475.831 Td /F1 9.8 Tf [(0.89)] TJ ET 0.267 0.267 0.267 rg 26.625 469.804 51.439 0.750 re f 26.625 453.923 0.750 16.631 re f 77.314 469.804 281.016 0.750 re f 77.314 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 459.949 Td /F1 9.8 Tf [(GFT \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 469.804 74.932 0.750 re f 357.580 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 459.949 Td /F1 9.8 Tf [(0.20)] TJ ET 0.267 0.267 0.267 rg 431.762 469.804 74.932 0.750 re f 431.762 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 459.949 Td /F1 9.8 Tf [(0.45)] TJ ET 0.267 0.267 0.267 rg 505.943 469.804 74.932 0.750 re f 505.943 453.923 0.750 16.631 re f 580.125 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 459.949 Td /F1 9.8 Tf [(0.28)] TJ ET 0.267 0.267 0.267 rg 26.625 453.923 51.439 0.750 re f 26.625 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 444.068 Td /F1 9.8 Tf [(\(e\))] TJ ET 0.267 0.267 0.267 rg 77.314 453.923 281.016 0.750 re f 77.314 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 444.068 Td /F1 9.8 Tf [(Twitter + GFT \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 453.923 74.932 0.750 re f 357.580 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 444.068 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 431.762 453.923 74.932 0.750 re f 431.762 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 444.068 Td /F1 9.8 Tf [(0.67)] TJ ET 0.267 0.267 0.267 rg 505.943 453.923 74.932 0.750 re f 505.943 438.042 0.750 16.631 re f 580.125 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 444.068 Td /F1 9.8 Tf [(0.62)] TJ ET 0.267 0.267 0.267 rg 26.625 438.042 51.439 0.750 re f 26.625 422.161 0.750 16.631 re f 77.314 438.042 281.016 0.750 re f 77.314 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 428.187 Td /F1 9.8 Tf [(Twitter + GFT \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 438.042 74.932 0.750 re f 357.580 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 428.187 Td /F1 9.8 Tf [(0.15)] TJ ET 0.267 0.267 0.267 rg 431.762 438.042 74.932 0.750 re f 431.762 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 428.187 Td /F1 9.8 Tf [(0.33)] TJ ET 0.267 0.267 0.267 rg 505.943 438.042 74.932 0.750 re f 505.943 422.161 0.750 16.631 re f 580.125 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 428.187 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 26.625 422.161 51.439 0.750 re f 26.625 406.279 51.439 0.750 re f 26.625 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 412.306 Td /F1 9.8 Tf [(\(f\))] TJ ET 0.267 0.267 0.267 rg 77.314 422.161 281.016 0.750 re f 77.314 406.279 281.016 0.750 re f 77.314 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 412.306 Td /F1 9.8 Tf [(Historical Average \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 422.161 74.932 0.750 re f 357.580 406.279 74.932 0.750 re f 357.580 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 412.306 Td /F1 9.8 Tf [(0.95)] TJ ET 0.267 0.267 0.267 rg 431.762 422.161 74.932 0.750 re f 431.762 406.279 74.932 0.750 re f 431.762 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 412.306 Td /F1 9.8 Tf [(0.87)] TJ ET 0.267 0.267 0.267 rg 505.943 422.161 74.932 0.750 re f 505.943 406.279 74.932 0.750 re f 505.943 406.279 0.750 16.631 re f 580.125 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 412.306 Td /F1 9.8 Tf [(1.39)] TJ ET BT 26.250 351.381 Td /F1 9.8 Tf [(We trained several linear autoregressive models on ILINet data from 2011-2013, and explored their ability to correctly forecast )] TJ ET BT 26.250 339.476 Td /F1 9.8 Tf [(the next weeks influenza rate \(k=0\). We found that a model incorporating Twitter data outperformed an equivalent model relying )] TJ ET BT 26.250 327.571 Td /F1 9.8 Tf [(only upon historical ILINet data \(see Figures 1-2\). In addition, Table 1 shows that Twitter improves forecasting in all three )] TJ ET BT 26.250 315.666 Td /F1 9.8 Tf [(seasons. In contrast, GFT failed to reduce error in two of the three seasons. Furthermore, adding GFT data to a model that )] TJ ET BT 26.250 303.762 Td /F1 9.8 Tf [(already incorporates ILINet and Twitter data actually reduces performance. \(GFTs worst season is 2012-13, likely due to its )] TJ ET BT 26.250 291.857 Td /F1 9.8 Tf [(gross overestimate of the peak influenza rate. )] TJ ET 0.267 0.267 0.267 rg BT 226.222 293.364 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 235.860 295.745 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 238.269 293.364 Td /F4 8.7 Tf [(21)] TJ ET 0.271 0.267 0.267 rg BT 247.907 291.857 Td /F1 9.8 Tf [( \) Finally, we find that errors using historically available ILINet data are, on )] TJ ET BT 26.250 279.952 Td /F1 9.8 Tf [(average, 42% higher as compared to CDCs revised estimates that were not available at the time of the forecast. Twitter )] TJ ET BT 26.250 268.047 Td /F1 9.8 Tf [(forecasts always improve upon those that only use historical data. Moreover, the reduction in error that Twitter provides is )] TJ ET BT 26.250 256.143 Td /F1 9.8 Tf [(substantially understated when using the CDCs revised estimates rather than the initially reported values. Incorporating Twitter )] TJ ET BT 26.250 244.238 Td /F1 9.8 Tf [(reduces nowcasting error by 29.6% when using the values available at the time of the nowcast, but only reduces error by 6.09% )] TJ ET BT 26.250 232.333 Td /F1 9.8 Tf [(when using the final estimates.)] TJ ET Q q 15.000 168.651 577.500 608.349 re W n 0.271 0.267 0.267 rg BT 26.250 767.476 Td /F1 9.8 Tf [(We include a nonparametric baseline, which predicts each weeks value as the average value from all historical data for that )] TJ ET BT 26.250 755.571 Td /F1 9.8 Tf [(week from 1997-2010. That is,)] TJ ET q 229.500 0 0 24.750 26.250 720.940 cm /I42 Do Q BT 26.250 703.917 Td /F1 9.8 Tf [(where y)] TJ ET BT 60.385 707.805 Td /F1 8.7 Tf [(i)] TJ ET BT 62.309 701.852 Td /F1 8.7 Tf [(w )] TJ ET BT 70.975 703.917 Td /F1 9.8 Tf [(is the value at week w in the season starting in year i and ending in year i+1.)] TJ ET BT 26.250 684.512 Td /F1 9.8 Tf [(The purpose of this comparison is to understand how much information we are gaining using autoregressive models over simply )] TJ ET BT 26.250 672.607 Td /F1 9.8 Tf [(modeling each season as the average of previous seasons.)] TJ ET BT 26.250 636.005 Td /F4 12.0 Tf [(Results)] TJ ET q 26.250 611.838 555.000 13.736 re W n 0.271 0.267 0.267 rg BT 26.250 614.585 Td /F1 9.8 Tf [(Table 1. Mean absolute errors from cross-validation across three seasons for the nowcasting task)] TJ ET Q 0.965 0.965 0.965 rg 26.250 562.615 555.000 41.722 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 604.338 m 581.250 604.338 l 580.500 603.588 l 27.000 603.588 l f 581.250 604.338 m 581.250 562.615 l 580.500 562.615 l 580.500 603.588 l f 26.250 604.338 m 26.250 562.615 l 27.000 562.615 l 27.000 603.588 l f 0.271 0.267 0.267 rg BT 33.000 590.262 Td /F1 9.0 Tf [(\(a\) Final revised CDC weekly estimates; \(b\) the realistic model using original CDC data before revision; \(c\) the model augmented with )] TJ ET BT 33.000 581.104 Td /F1 9.0 Tf [(Twitter data; \(d\) the model augmented with GFT data; \(e\) the model augmented with both Twitter and GFT data; \(f\) values predicted by )] TJ ET BT 33.000 571.947 Td /F1 9.0 Tf [(measuring the historical average.)] TJ ET 1.000 1.000 1.000 rg 26.250 405.904 555.000 156.711 re f 0.965 0.965 0.965 rg 27.000 549.586 50.689 12.280 re f 0.267 0.267 0.267 rg 26.625 561.490 51.064 0.750 re f 26.625 549.211 0.750 13.030 re f 0.965 0.965 0.965 rg 77.689 549.586 280.266 12.280 re f 0.267 0.267 0.267 rg 77.689 561.490 280.266 0.750 re f 0.271 0.267 0.267 rg BT 82.189 552.342 Td /F4 9.8 Tf [(Model)] TJ ET 0.965 0.965 0.965 rg 357.955 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 357.955 561.490 74.182 0.750 re f 0.271 0.267 0.267 rg BT 362.455 552.342 Td /F4 9.8 Tf [(11-12)] TJ ET 0.965 0.965 0.965 rg 432.137 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 432.137 561.490 74.182 0.750 re f 0.271 0.267 0.267 rg BT 436.637 552.342 Td /F4 9.8 Tf [(12-13)] TJ ET 0.965 0.965 0.965 rg 506.318 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 506.318 561.490 74.557 0.750 re f 580.125 549.211 0.750 13.030 re f 0.271 0.267 0.267 rg BT 510.818 552.342 Td /F4 9.8 Tf [(13-14)] TJ ET 0.267 0.267 0.267 rg 26.625 549.211 51.439 0.750 re f 26.625 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 539.356 Td /F1 9.8 Tf [(\(a\))] TJ ET 0.267 0.267 0.267 rg 77.314 549.211 281.016 0.750 re f 77.314 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 539.356 Td /F1 9.8 Tf [(Revised CDC \(y\))] TJ ET 0.267 0.267 0.267 rg 357.580 549.211 74.932 0.750 re f 357.580 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 539.356 Td /F1 9.8 Tf [(0.10)] TJ ET 0.267 0.267 0.267 rg 431.762 549.211 74.932 0.750 re f 431.762 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 539.356 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 505.943 549.211 74.932 0.750 re f 505.943 533.329 0.750 16.631 re f 580.125 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 539.356 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 26.625 533.329 51.439 0.750 re f 26.625 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 523.474 Td /F1 9.8 Tf [(\(b\))] TJ ET 0.267 0.267 0.267 rg 77.314 533.329 281.016 0.750 re f 77.314 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 523.474 Td /F1 9.8 Tf [(Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 533.329 74.932 0.750 re f 357.580 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 523.474 Td /F1 9.8 Tf [(0.20)] TJ ET 0.267 0.267 0.267 rg 431.762 533.329 74.932 0.750 re f 431.762 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 523.474 Td /F1 9.8 Tf [(0.30)] TJ ET 0.267 0.267 0.267 rg 505.943 533.329 74.932 0.750 re f 505.943 517.448 0.750 16.631 re f 580.125 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 523.474 Td /F1 9.8 Tf [(0.32)] TJ ET 0.267 0.267 0.267 rg 26.625 517.448 51.439 0.750 re f 26.625 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 507.593 Td /F1 9.8 Tf [(\(c\))] TJ ET 0.267 0.267 0.267 rg 77.314 517.448 281.016 0.750 re f 77.314 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 507.593 Td /F1 9.8 Tf [(Twitter \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 517.448 74.932 0.750 re f 357.580 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 507.593 Td /F1 9.8 Tf [(0.33)] TJ ET 0.267 0.267 0.267 rg 431.762 517.448 74.932 0.750 re f 431.762 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 507.593 Td /F1 9.8 Tf [(0.36)] TJ ET 0.267 0.267 0.267 rg 505.943 517.448 74.932 0.750 re f 505.943 501.567 0.750 16.631 re f 580.125 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 507.593 Td /F1 9.8 Tf [(0.48)] TJ ET 0.267 0.267 0.267 rg 26.625 501.567 51.439 0.750 re f 26.625 485.686 0.750 16.631 re f 77.314 501.567 281.016 0.750 re f 77.314 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 491.712 Td /F1 9.8 Tf [(Twitter \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 501.567 74.932 0.750 re f 357.580 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 491.712 Td /F1 9.8 Tf [(0.14)] TJ ET 0.267 0.267 0.267 rg 431.762 501.567 74.932 0.750 re f 431.762 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 491.712 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 505.943 501.567 74.932 0.750 re f 505.943 485.686 0.750 16.631 re f 580.125 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 491.712 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 26.625 485.686 51.439 0.750 re f 26.625 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 475.831 Td /F1 9.8 Tf [(\(d\))] TJ ET 0.267 0.267 0.267 rg 77.314 485.686 281.016 0.750 re f 77.314 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 475.831 Td /F1 9.8 Tf [(GFT \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 485.686 74.932 0.750 re f 357.580 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 475.831 Td /F1 9.8 Tf [(0.35)] TJ ET 0.267 0.267 0.267 rg 431.762 485.686 74.932 0.750 re f 431.762 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 475.831 Td /F1 9.8 Tf [(0.71)] TJ ET 0.267 0.267 0.267 rg 505.943 485.686 74.932 0.750 re f 505.943 469.804 0.750 16.631 re f 580.125 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 475.831 Td /F1 9.8 Tf [(0.89)] TJ ET 0.267 0.267 0.267 rg 26.625 469.804 51.439 0.750 re f 26.625 453.923 0.750 16.631 re f 77.314 469.804 281.016 0.750 re f 77.314 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 459.949 Td /F1 9.8 Tf [(GFT \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 469.804 74.932 0.750 re f 357.580 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 459.949 Td /F1 9.8 Tf [(0.20)] TJ ET 0.267 0.267 0.267 rg 431.762 469.804 74.932 0.750 re f 431.762 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 459.949 Td /F1 9.8 Tf [(0.45)] TJ ET 0.267 0.267 0.267 rg 505.943 469.804 74.932 0.750 re f 505.943 453.923 0.750 16.631 re f 580.125 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 459.949 Td /F1 9.8 Tf [(0.28)] TJ ET 0.267 0.267 0.267 rg 26.625 453.923 51.439 0.750 re f 26.625 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 444.068 Td /F1 9.8 Tf [(\(e\))] TJ ET 0.267 0.267 0.267 rg 77.314 453.923 281.016 0.750 re f 77.314 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 444.068 Td /F1 9.8 Tf [(Twitter + GFT \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 453.923 74.932 0.750 re f 357.580 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 444.068 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 431.762 453.923 74.932 0.750 re f 431.762 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 444.068 Td /F1 9.8 Tf [(0.67)] TJ ET 0.267 0.267 0.267 rg 505.943 453.923 74.932 0.750 re f 505.943 438.042 0.750 16.631 re f 580.125 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 444.068 Td /F1 9.8 Tf [(0.62)] TJ ET 0.267 0.267 0.267 rg 26.625 438.042 51.439 0.750 re f 26.625 422.161 0.750 16.631 re f 77.314 438.042 281.016 0.750 re f 77.314 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 428.187 Td /F1 9.8 Tf [(Twitter + GFT \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 438.042 74.932 0.750 re f 357.580 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 428.187 Td /F1 9.8 Tf [(0.15)] TJ ET 0.267 0.267 0.267 rg 431.762 438.042 74.932 0.750 re f 431.762 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 428.187 Td /F1 9.8 Tf [(0.33)] TJ ET 0.267 0.267 0.267 rg 505.943 438.042 74.932 0.750 re f 505.943 422.161 0.750 16.631 re f 580.125 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 428.187 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 26.625 422.161 51.439 0.750 re f 26.625 406.279 51.439 0.750 re f 26.625 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 412.306 Td /F1 9.8 Tf [(\(f\))] TJ ET 0.267 0.267 0.267 rg 77.314 422.161 281.016 0.750 re f 77.314 406.279 281.016 0.750 re f 77.314 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 412.306 Td /F1 9.8 Tf [(Historical Average \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 422.161 74.932 0.750 re f 357.580 406.279 74.932 0.750 re f 357.580 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 412.306 Td /F1 9.8 Tf [(0.95)] TJ ET 0.267 0.267 0.267 rg 431.762 422.161 74.932 0.750 re f 431.762 406.279 74.932 0.750 re f 431.762 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 412.306 Td /F1 9.8 Tf [(0.87)] TJ ET 0.267 0.267 0.267 rg 505.943 422.161 74.932 0.750 re f 505.943 406.279 74.932 0.750 re f 505.943 406.279 0.750 16.631 re f 580.125 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 412.306 Td /F1 9.8 Tf [(1.39)] TJ ET BT 26.250 351.381 Td /F1 9.8 Tf [(We trained several linear autoregressive models on ILINet data from 2011-2013, and explored their ability to correctly forecast )] TJ ET BT 26.250 339.476 Td /F1 9.8 Tf [(the next weeks influenza rate \(k=0\). We found that a model incorporating Twitter data outperformed an equivalent model relying )] TJ ET BT 26.250 327.571 Td /F1 9.8 Tf [(only upon historical ILINet data \(see Figures 1-2\). In addition, Table 1 shows that Twitter improves forecasting in all three )] TJ ET BT 26.250 315.666 Td /F1 9.8 Tf [(seasons. In contrast, GFT failed to reduce error in two of the three seasons. Furthermore, adding GFT data to a model that )] TJ ET BT 26.250 303.762 Td /F1 9.8 Tf [(already incorporates ILINet and Twitter data actually reduces performance. \(GFTs worst season is 2012-13, likely due to its )] TJ ET BT 26.250 291.857 Td /F1 9.8 Tf [(gross overestimate of the peak influenza rate. )] TJ ET 0.267 0.267 0.267 rg BT 226.222 293.364 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 235.860 295.745 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 238.269 293.364 Td /F4 8.7 Tf [(21)] TJ ET 0.271 0.267 0.267 rg BT 247.907 291.857 Td /F1 9.8 Tf [( \) Finally, we find that errors using historically available ILINet data are, on )] TJ ET BT 26.250 279.952 Td /F1 9.8 Tf [(average, 42% higher as compared to CDCs revised estimates that were not available at the time of the forecast. Twitter )] TJ ET BT 26.250 268.047 Td /F1 9.8 Tf [(forecasts always improve upon those that only use historical data. Moreover, the reduction in error that Twitter provides is )] TJ ET BT 26.250 256.143 Td /F1 9.8 Tf [(substantially understated when using the CDCs revised estimates rather than the initially reported values. Incorporating Twitter )] TJ ET BT 26.250 244.238 Td /F1 9.8 Tf [(reduces nowcasting error by 29.6% when using the values available at the time of the nowcast, but only reduces error by 6.09% )] TJ ET BT 26.250 232.333 Td /F1 9.8 Tf [(when using the final estimates.)] TJ ET Q q 15.000 168.651 577.500 608.349 re W n 0.271 0.267 0.267 rg BT 26.250 767.476 Td /F1 9.8 Tf [(We include a nonparametric baseline, which predicts each weeks value as the average value from all historical data for that )] TJ ET BT 26.250 755.571 Td /F1 9.8 Tf [(week from 1997-2010. That is,)] TJ ET q 229.500 0 0 24.750 26.250 720.940 cm /I44 Do Q BT 26.250 703.917 Td /F1 9.8 Tf [(where y)] TJ ET BT 60.385 707.805 Td /F1 8.7 Tf [(i)] TJ ET BT 62.309 701.852 Td /F1 8.7 Tf [(w )] TJ ET BT 70.975 703.917 Td /F1 9.8 Tf [(is the value at week w in the season starting in year i and ending in year i+1.)] TJ ET BT 26.250 684.512 Td /F1 9.8 Tf [(The purpose of this comparison is to understand how much information we are gaining using autoregressive models over simply )] TJ ET BT 26.250 672.607 Td /F1 9.8 Tf [(modeling each season as the average of previous seasons.)] TJ ET BT 26.250 636.005 Td /F4 12.0 Tf [(Results)] TJ ET q 26.250 611.838 555.000 13.736 re W n 0.271 0.267 0.267 rg BT 26.250 614.585 Td /F1 9.8 Tf [(Table 1. Mean absolute errors from cross-validation across three seasons for the nowcasting task)] TJ ET Q 0.965 0.965 0.965 rg 26.250 562.615 555.000 41.722 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 604.338 m 581.250 604.338 l 580.500 603.588 l 27.000 603.588 l f 581.250 604.338 m 581.250 562.615 l 580.500 562.615 l 580.500 603.588 l f 26.250 604.338 m 26.250 562.615 l 27.000 562.615 l 27.000 603.588 l f 0.271 0.267 0.267 rg BT 33.000 590.262 Td /F1 9.0 Tf [(\(a\) Final revised CDC weekly estimates; \(b\) the realistic model using original CDC data before revision; \(c\) the model augmented with )] TJ ET BT 33.000 581.104 Td /F1 9.0 Tf [(Twitter data; \(d\) the model augmented with GFT data; \(e\) the model augmented with both Twitter and GFT data; \(f\) values predicted by )] TJ ET BT 33.000 571.947 Td /F1 9.0 Tf [(measuring the historical average.)] TJ ET 1.000 1.000 1.000 rg 26.250 405.904 555.000 156.711 re f 0.965 0.965 0.965 rg 27.000 549.586 50.689 12.280 re f 0.267 0.267 0.267 rg 26.625 561.490 51.064 0.750 re f 26.625 549.211 0.750 13.030 re f 0.965 0.965 0.965 rg 77.689 549.586 280.266 12.280 re f 0.267 0.267 0.267 rg 77.689 561.490 280.266 0.750 re f 0.271 0.267 0.267 rg BT 82.189 552.342 Td /F4 9.8 Tf [(Model)] TJ ET 0.965 0.965 0.965 rg 357.955 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 357.955 561.490 74.182 0.750 re f 0.271 0.267 0.267 rg BT 362.455 552.342 Td /F4 9.8 Tf [(11-12)] TJ ET 0.965 0.965 0.965 rg 432.137 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 432.137 561.490 74.182 0.750 re f 0.271 0.267 0.267 rg BT 436.637 552.342 Td /F4 9.8 Tf [(12-13)] TJ ET 0.965 0.965 0.965 rg 506.318 549.586 74.182 12.280 re f 0.267 0.267 0.267 rg 506.318 561.490 74.557 0.750 re f 580.125 549.211 0.750 13.030 re f 0.271 0.267 0.267 rg BT 510.818 552.342 Td /F4 9.8 Tf [(13-14)] TJ ET 0.267 0.267 0.267 rg 26.625 549.211 51.439 0.750 re f 26.625 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 539.356 Td /F1 9.8 Tf [(\(a\))] TJ ET 0.267 0.267 0.267 rg 77.314 549.211 281.016 0.750 re f 77.314 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 539.356 Td /F1 9.8 Tf [(Revised CDC \(y\))] TJ ET 0.267 0.267 0.267 rg 357.580 549.211 74.932 0.750 re f 357.580 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 539.356 Td /F1 9.8 Tf [(0.10)] TJ ET 0.267 0.267 0.267 rg 431.762 549.211 74.932 0.750 re f 431.762 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 539.356 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 505.943 549.211 74.932 0.750 re f 505.943 533.329 0.750 16.631 re f 580.125 533.329 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 539.356 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 26.625 533.329 51.439 0.750 re f 26.625 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 523.474 Td /F1 9.8 Tf [(\(b\))] TJ ET 0.267 0.267 0.267 rg 77.314 533.329 281.016 0.750 re f 77.314 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 523.474 Td /F1 9.8 Tf [(Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 533.329 74.932 0.750 re f 357.580 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 523.474 Td /F1 9.8 Tf [(0.20)] TJ ET 0.267 0.267 0.267 rg 431.762 533.329 74.932 0.750 re f 431.762 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 523.474 Td /F1 9.8 Tf [(0.30)] TJ ET 0.267 0.267 0.267 rg 505.943 533.329 74.932 0.750 re f 505.943 517.448 0.750 16.631 re f 580.125 517.448 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 523.474 Td /F1 9.8 Tf [(0.32)] TJ ET 0.267 0.267 0.267 rg 26.625 517.448 51.439 0.750 re f 26.625 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 507.593 Td /F1 9.8 Tf [(\(c\))] TJ ET 0.267 0.267 0.267 rg 77.314 517.448 281.016 0.750 re f 77.314 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 507.593 Td /F1 9.8 Tf [(Twitter \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 517.448 74.932 0.750 re f 357.580 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 507.593 Td /F1 9.8 Tf [(0.33)] TJ ET 0.267 0.267 0.267 rg 431.762 517.448 74.932 0.750 re f 431.762 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 507.593 Td /F1 9.8 Tf [(0.36)] TJ ET 0.267 0.267 0.267 rg 505.943 517.448 74.932 0.750 re f 505.943 501.567 0.750 16.631 re f 580.125 501.567 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 507.593 Td /F1 9.8 Tf [(0.48)] TJ ET 0.267 0.267 0.267 rg 26.625 501.567 51.439 0.750 re f 26.625 485.686 0.750 16.631 re f 77.314 501.567 281.016 0.750 re f 77.314 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 491.712 Td /F1 9.8 Tf [(Twitter \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 501.567 74.932 0.750 re f 357.580 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 491.712 Td /F1 9.8 Tf [(0.14)] TJ ET 0.267 0.267 0.267 rg 431.762 501.567 74.932 0.750 re f 431.762 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 491.712 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 505.943 501.567 74.932 0.750 re f 505.943 485.686 0.750 16.631 re f 580.125 485.686 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 491.712 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 26.625 485.686 51.439 0.750 re f 26.625 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 475.831 Td /F1 9.8 Tf [(\(d\))] TJ ET 0.267 0.267 0.267 rg 77.314 485.686 281.016 0.750 re f 77.314 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 475.831 Td /F1 9.8 Tf [(GFT \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 485.686 74.932 0.750 re f 357.580 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 475.831 Td /F1 9.8 Tf [(0.35)] TJ ET 0.267 0.267 0.267 rg 431.762 485.686 74.932 0.750 re f 431.762 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 475.831 Td /F1 9.8 Tf [(0.71)] TJ ET 0.267 0.267 0.267 rg 505.943 485.686 74.932 0.750 re f 505.943 469.804 0.750 16.631 re f 580.125 469.804 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 475.831 Td /F1 9.8 Tf [(0.89)] TJ ET 0.267 0.267 0.267 rg 26.625 469.804 51.439 0.750 re f 26.625 453.923 0.750 16.631 re f 77.314 469.804 281.016 0.750 re f 77.314 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 459.949 Td /F1 9.8 Tf [(GFT \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 469.804 74.932 0.750 re f 357.580 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 459.949 Td /F1 9.8 Tf [(0.20)] TJ ET 0.267 0.267 0.267 rg 431.762 469.804 74.932 0.750 re f 431.762 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 459.949 Td /F1 9.8 Tf [(0.45)] TJ ET 0.267 0.267 0.267 rg 505.943 469.804 74.932 0.750 re f 505.943 453.923 0.750 16.631 re f 580.125 453.923 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 459.949 Td /F1 9.8 Tf [(0.28)] TJ ET 0.267 0.267 0.267 rg 26.625 453.923 51.439 0.750 re f 26.625 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 444.068 Td /F1 9.8 Tf [(\(e\))] TJ ET 0.267 0.267 0.267 rg 77.314 453.923 281.016 0.750 re f 77.314 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 444.068 Td /F1 9.8 Tf [(Twitter + GFT \(z\))] TJ ET 0.267 0.267 0.267 rg 357.580 453.923 74.932 0.750 re f 357.580 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 444.068 Td /F1 9.8 Tf [(0.24)] TJ ET 0.267 0.267 0.267 rg 431.762 453.923 74.932 0.750 re f 431.762 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 444.068 Td /F1 9.8 Tf [(0.67)] TJ ET 0.267 0.267 0.267 rg 505.943 453.923 74.932 0.750 re f 505.943 438.042 0.750 16.631 re f 580.125 438.042 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 444.068 Td /F1 9.8 Tf [(0.62)] TJ ET 0.267 0.267 0.267 rg 26.625 438.042 51.439 0.750 re f 26.625 422.161 0.750 16.631 re f 77.314 438.042 281.016 0.750 re f 77.314 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 428.187 Td /F1 9.8 Tf [(Twitter + GFT \(z\) + Original CDC \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 438.042 74.932 0.750 re f 357.580 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 428.187 Td /F1 9.8 Tf [(0.15)] TJ ET 0.267 0.267 0.267 rg 431.762 438.042 74.932 0.750 re f 431.762 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 428.187 Td /F1 9.8 Tf [(0.33)] TJ ET 0.267 0.267 0.267 rg 505.943 438.042 74.932 0.750 re f 505.943 422.161 0.750 16.631 re f 580.125 422.161 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 428.187 Td /F1 9.8 Tf [(0.21)] TJ ET 0.267 0.267 0.267 rg 26.625 422.161 51.439 0.750 re f 26.625 406.279 51.439 0.750 re f 26.625 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 412.306 Td /F1 9.8 Tf [(\(f\))] TJ ET 0.267 0.267 0.267 rg 77.314 422.161 281.016 0.750 re f 77.314 406.279 281.016 0.750 re f 77.314 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 82.564 412.306 Td /F1 9.8 Tf [(Historical Average \(~y\))] TJ ET 0.267 0.267 0.267 rg 357.580 422.161 74.932 0.750 re f 357.580 406.279 74.932 0.750 re f 357.580 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 362.830 412.306 Td /F1 9.8 Tf [(0.95)] TJ ET 0.267 0.267 0.267 rg 431.762 422.161 74.932 0.750 re f 431.762 406.279 74.932 0.750 re f 431.762 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 437.012 412.306 Td /F1 9.8 Tf [(0.87)] TJ ET 0.267 0.267 0.267 rg 505.943 422.161 74.932 0.750 re f 505.943 406.279 74.932 0.750 re f 505.943 406.279 0.750 16.631 re f 580.125 406.279 0.750 16.631 re f 0.271 0.267 0.267 rg BT 511.193 412.306 Td /F1 9.8 Tf [(1.39)] TJ ET BT 26.250 351.381 Td /F1 9.8 Tf [(We trained several linear autoregressive models on ILINet data from 2011-2013, and explored their ability to correctly forecast )] TJ ET BT 26.250 339.476 Td /F1 9.8 Tf [(the next weeks influenza rate \(k=0\). We found that a model incorporating Twitter data outperformed an equivalent model relying )] TJ ET BT 26.250 327.571 Td /F1 9.8 Tf [(only upon historical ILINet data \(see Figures 1-2\). In addition, Table 1 shows that Twitter improves forecasting in all three )] TJ ET BT 26.250 315.666 Td /F1 9.8 Tf [(seasons. In contrast, GFT failed to reduce error in two of the three seasons. Furthermore, adding GFT data to a model that )] TJ ET BT 26.250 303.762 Td /F1 9.8 Tf [(already incorporates ILINet and Twitter data actually reduces performance. \(GFTs worst season is 2012-13, likely due to its )] TJ ET BT 26.250 291.857 Td /F1 9.8 Tf [(gross overestimate of the peak influenza rate. )] TJ ET 0.267 0.267 0.267 rg BT 226.222 293.364 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 235.860 295.745 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 238.269 293.364 Td /F4 8.7 Tf [(21)] TJ ET 0.271 0.267 0.267 rg BT 247.907 291.857 Td /F1 9.8 Tf [( \) Finally, we find that errors using historically available ILINet data are, on )] TJ ET BT 26.250 279.952 Td /F1 9.8 Tf [(average, 42% higher as compared to CDCs revised estimates that were not available at the time of the forecast. Twitter )] TJ ET BT 26.250 268.047 Td /F1 9.8 Tf [(forecasts always improve upon those that only use historical data. Moreover, the reduction in error that Twitter provides is )] TJ ET BT 26.250 256.143 Td /F1 9.8 Tf [(substantially understated when using the CDCs revised estimates rather than the initially reported values. Incorporating Twitter )] TJ ET BT 26.250 244.238 Td /F1 9.8 Tf [(reduces nowcasting error by 29.6% when using the values available at the time of the nowcast, but only reduces error by 6.09% )] TJ ET BT 26.250 232.333 Td /F1 9.8 Tf [(when using the final estimates.)] TJ ET Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(3)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj 337 0 obj << /Type /XObject /Subtype /Image /Width 306 /Height 33 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 306 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1735>> stream hYg#[:GQybsUE}XתU֊Q뺪F*Fb>TDjĊyXQ?~3nI4{{jϙ)DlS'ON){`5u\r(#K[X5GG(l?MYۡmSo6cʃtwҝ {4 ebmyHyXLBBτw肎c6O+'? )j~8[lnyM)1ˡP4 ͹?b@ɳtxvڟט@WAtX4l+@)2X&k{ LKd MޟRR)s;Tʪ87s7潦(+,x_GwnH"Z5,V`]L jk3^z hPU I8:D5u v93IQILD&uFÉt[VGi2ݡ9:&s *tݶ2xxӧva6eݐI ]ڜ71tzxTOfΑ&" ђj=2;bUFB@+V1a#̙u\m1LMIK,cS%g$_\ ^09Qr XuKA'_qn;Nl>gyνu)Cdsj GUt L娆BTaF~3Oe|y&PfmcX|/uR2&}'c!ʔ_dO|V*M%:7h}:K%]u=*6Sj%4G_?f6qkZUcͧ?д_*$ iKz>}7?!b8ȯǜ궚ߒV7 endstream endobj 338 0 obj << /Type /XObject /Subtype /Image /Width 306 /Height 33 /SMask 337 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 306 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 52>> stream x1 Om 82vw endstream endobj 339 0 obj << /Type /Annot /Subtype /Link /A 340 0 R /Border [0 0 0] /H /I /Rect [ 226.2225 292.5624 235.8598 301.3807 ] >> endobj 340 0 obj << /Type /Action >> endobj 341 0 obj << /Type /Annot /Subtype /Link /A 342 0 R /Border [0 0 0] /H /I /Rect [ 238.2692 292.5624 247.9065 301.3807 ] >> endobj 342 0 obj << /Type /Action >> endobj 343 0 obj << /Type /XObject /Subtype /Image /Width 306 /Height 33 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 306 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1735>> stream hYg#[:GQybsUE}XתU֊Q뺪F*Fb>TDjĊyXQ?~3nI4{{jϙ)DlS'ON){`5u\r(#K[X5GG(l?MYۡmSo6cʃtwҝ {4 ebmyHyXLBBτw肎c6O+'? )j~8[lnyM)1ˡP4 ͹?b@ɳtxvڟט@WAtX4l+@)2X&k{ LKd MޟRR)s;Tʪ87s7潦(+,x_GwnH"Z5,V`]L jk3^z hPU I8:D5u v93IQILD&uFÉt[VGi2ݡ9:&s *tݶ2xxӧva6eݐI ]ڜ71tzxTOfΑ&" ђj=2;bUFB@+V1a#̙u\m1LMIK,cS%g$_\ ^09Qr XuKA'_qn;Nl>gyνu)Cdsj GUt L娆BTaF~3Oe|y&PfmcX|/uR2&}'c!ʔ_dO|V*M%:7h}:K%]u=*6Sj%4G_?f6qkZUcͧ?д_*$ iKz>}7?!b8ȯǜ궚ߒV7 endstream endobj 344 0 obj << /Type /XObject /Subtype /Image /Width 306 /Height 33 /SMask 343 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 306 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 52>> stream x1 Om 82vw endstream endobj 345 0 obj << /Type /Annot /Subtype /Link /A 346 0 R /Border [0 0 0] /H /I /Rect [ 226.2225 292.5624 235.8598 301.3807 ] >> endobj 346 0 obj << /Type /Action >> endobj 347 0 obj << /Type /Annot /Subtype /Link /A 348 0 R /Border [0 0 0] /H /I /Rect [ 238.2692 292.5624 247.9065 301.3807 ] >> endobj 348 0 obj << /Type /Action >> endobj 349 0 obj << /Type /XObject /Subtype /Image /Width 306 /Height 33 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 306 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 1735>> stream hYg#[:GQybsUE}XתU֊Q뺪F*Fb>TDjĊyXQ?~3nI4{{jϙ)DlS'ON){`5u\r(#K[X5GG(l?MYۡmSo6cʃtwҝ {4 ebmyHyXLBBτw肎c6O+'? )j~8[lnyM)1ˡP4 ͹?b@ɳtxvڟט@WAtX4l+@)2X&k{ LKd MޟRR)s;Tʪ87s7潦(+,x_GwnH"Z5,V`]L jk3^z hPU I8:D5u v93IQILD&uFÉt[VGi2ݡ9:&s *tݶ2xxӧva6eݐI ]ڜ71tzxTOfΑ&" ђj=2;bUFB@+V1a#̙u\m1LMIK,cS%g$_\ ^09Qr XuKA'_qn;Nl>gyνu)Cdsj GUt L娆BTaF~3Oe|y&PfmcX|/uR2&}'c!ʔ_dO|V*M%:7h}:K%]u=*6Sj%4G_?f6qkZUcͧ?д_*$ iKz>}7?!b8ȯǜ궚ߒV7 endstream endobj 350 0 obj << /Type /XObject /Subtype /Image /Width 306 /Height 33 /SMask 349 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 306 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 52>> stream x1 Om 82vw endstream endobj 351 0 obj << /Type /Annot /Subtype /Link /A 352 0 R /Border [0 0 0] /H /I /Rect [ 226.2225 292.5624 235.8598 301.3807 ] >> endobj 352 0 obj << /Type /Action >> endobj 353 0 obj << /Type /Annot /Subtype /Link /A 354 0 R /Border [0 0 0] /H /I /Rect [ 238.2692 292.5624 247.9065 301.3807 ] >> endobj 354 0 obj << /Type /Action >> endobj 355 0 obj << /Type /Page /Parent 3 0 R /Contents 356 0 R >> endobj 356 0 obj << /Length 31546 >> stream 0.271 0.267 0.267 rg 0.267 0.267 0.267 RG q 15.000 169.742 577.500 607.258 re W n q 26.250 763.264 555.000 13.736 re W n 0.271 0.267 0.267 rg BT 26.250 766.011 Td /F1 9.8 Tf [(Table 2. The mean \(+/- SD\) absolute error of two forecasting models after k weeks.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 723.199 555.000 32.565 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 755.764 m 581.250 755.764 l 580.500 755.014 l 27.000 755.014 l f 581.250 755.764 m 581.250 723.199 l 580.500 723.199 l 580.500 755.014 l f 26.250 755.764 m 26.250 723.199 l 27.000 723.199 l 27.000 755.014 l f 0.271 0.267 0.267 rg BT 33.000 741.688 Td /F1 9.0 Tf [(We compared the baseline model based on previous weeks of CDC ILI data to the best-performing model, that incorporating Twitter )] TJ ET BT 33.000 732.530 Td /F1 9.0 Tf [(data. The third column shows the error reduction over the baseline when using the Twitter model.)] TJ ET 1.000 1.000 1.000 rg 26.250 373.113 555.000 350.086 re f 0.965 0.965 0.965 rg 27.000 710.169 58.756 12.280 re f 0.267 0.267 0.267 rg 26.625 722.074 59.131 0.750 re f 26.625 709.794 0.750 13.030 re f 0.271 0.267 0.267 rg BT 31.500 712.925 Td /F4 9.8 Tf [(k)] TJ ET 0.965 0.965 0.965 rg 85.756 710.169 139.673 12.280 re f 0.267 0.267 0.267 rg 85.756 722.074 139.673 0.750 re f 0.271 0.267 0.267 rg BT 90.256 712.925 Td /F4 9.8 Tf [(CDC Only)] TJ ET 0.965 0.965 0.965 rg 225.429 710.169 193.379 12.280 re f 0.267 0.267 0.267 rg 225.429 722.074 193.379 0.750 re f 0.271 0.267 0.267 rg BT 229.929 712.925 Td /F4 9.8 Tf [(CDC+Twitter)] TJ ET 0.965 0.965 0.965 rg 418.808 710.169 161.692 12.280 re f 0.267 0.267 0.267 rg 418.808 722.074 162.067 0.750 re f 580.125 709.794 0.750 13.030 re f 0.271 0.267 0.267 rg BT 423.308 712.925 Td /F4 9.8 Tf [(Reduction)] TJ ET 0.267 0.267 0.267 rg 26.625 709.794 59.506 0.750 re f 26.625 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 699.939 Td /F1 9.8 Tf [(0)] TJ ET 0.267 0.267 0.267 rg 85.381 709.794 140.423 0.750 re f 85.381 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 699.939 Td /F1 9.8 Tf [(0.27 0.06)] TJ ET 0.267 0.267 0.267 rg 225.054 709.794 194.129 0.750 re f 225.054 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 699.939 Td /F1 9.8 Tf [(0.19 0.03)] TJ ET 0.267 0.267 0.267 rg 418.433 709.794 162.442 0.750 re f 418.433 693.913 0.750 16.631 re f 580.125 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 699.939 Td /F1 9.8 Tf [(29.6%)] TJ ET 0.267 0.267 0.267 rg 26.625 693.913 59.506 0.750 re f 26.625 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 684.058 Td /F1 9.8 Tf [(1)] TJ ET 0.267 0.267 0.267 rg 85.381 693.913 140.423 0.750 re f 85.381 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 684.058 Td /F1 9.8 Tf [(0.40 0.12)] TJ ET 0.267 0.267 0.267 rg 225.054 693.913 194.129 0.750 re f 225.054 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 684.058 Td /F1 9.8 Tf [(0.29 0.07)] TJ ET 0.267 0.267 0.267 rg 418.433 693.913 162.442 0.750 re f 418.433 678.031 0.750 16.631 re f 580.125 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 684.058 Td /F1 9.8 Tf [(27.5%)] TJ ET 0.267 0.267 0.267 rg 26.625 678.031 59.506 0.750 re f 26.625 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 668.176 Td /F1 9.8 Tf [(2)] TJ ET 0.267 0.267 0.267 rg 85.381 678.031 140.423 0.750 re f 85.381 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 668.176 Td /F1 9.8 Tf [(0.49 0.17)] TJ ET 0.267 0.267 0.267 rg 225.054 678.031 194.129 0.750 re f 225.054 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 668.176 Td /F1 9.8 Tf [(0.37 0.08)] TJ ET 0.267 0.267 0.267 rg 418.433 678.031 162.442 0.750 re f 418.433 662.150 0.750 16.631 re f 580.125 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 668.176 Td /F1 9.8 Tf [(24.5%)] TJ ET 0.267 0.267 0.267 rg 26.625 662.150 59.506 0.750 re f 26.625 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 652.295 Td /F1 9.8 Tf [(3)] TJ ET 0.267 0.267 0.267 rg 85.381 662.150 140.423 0.750 re f 85.381 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 652.295 Td /F1 9.8 Tf [(0.59 0.22)] TJ ET 0.267 0.267 0.267 rg 225.054 662.150 194.129 0.750 re f 225.054 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 652.295 Td /F1 9.8 Tf [(0.46 0.11)] TJ ET 0.267 0.267 0.267 rg 418.433 662.150 162.442 0.750 re f 418.433 646.269 0.750 16.631 re f 580.125 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 652.295 Td /F1 9.8 Tf [(22.0%)] TJ ET 0.267 0.267 0.267 rg 26.625 646.269 59.506 0.750 re f 26.625 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 636.414 Td /F1 9.8 Tf [(4)] TJ ET 0.267 0.267 0.267 rg 85.381 646.269 140.423 0.750 re f 85.381 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 636.414 Td /F1 9.8 Tf [(0.72 0.27)] TJ ET 0.267 0.267 0.267 rg 225.054 646.269 194.129 0.750 re f 225.054 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 636.414 Td /F1 9.8 Tf [(0.55 0.14)] TJ ET 0.267 0.267 0.267 rg 418.433 646.269 162.442 0.750 re f 418.433 630.388 0.750 16.631 re f 580.125 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 636.414 Td /F1 9.8 Tf [(25.3%)] TJ ET 0.267 0.267 0.267 rg 26.625 630.388 59.506 0.750 re f 26.625 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 31.875 620.533 Td /F1 9.8 Tf [(5)] TJ ET 0.267 0.267 0.267 rg 85.381 630.388 140.423 0.750 re f 85.381 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 90.631 620.533 Td /F1 9.8 Tf [(0.83 0.33)] TJ ET 0.267 0.267 0.267 rg 225.054 630.388 194.129 0.750 re f 225.054 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 230.304 620.533 Td /F1 9.8 Tf [(0.64 0.17)] TJ ET 0.267 0.267 0.267 rg 418.433 630.388 162.442 0.750 re f 418.433 452.894 0.750 178.244 re f 580.125 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 423.683 620.533 Td /F1 9.8 Tf [(23.6%)] TJ ET 0.267 0.267 0.267 rg 26.625 452.894 59.506 0.750 re f 26.625 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 443.039 Td /F1 9.8 Tf [(6)] TJ ET 0.267 0.267 0.267 rg 85.381 452.894 140.423 0.750 re f 85.381 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 443.039 Td /F1 9.8 Tf [(0.92 0.39)] TJ ET 0.267 0.267 0.267 rg 225.054 452.894 194.129 0.750 re f 225.054 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 443.039 Td /F1 9.8 Tf [(0.73 0.21)] TJ ET 0.267 0.267 0.267 rg 418.433 452.894 162.442 0.750 re f 418.433 437.013 0.750 16.631 re f 580.125 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 443.039 Td /F1 9.8 Tf [(20.7%)] TJ ET 0.267 0.267 0.267 rg 26.625 437.013 59.506 0.750 re f 26.625 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 427.158 Td /F1 9.8 Tf [(7)] TJ ET 0.267 0.267 0.267 rg 85.381 437.013 140.423 0.750 re f 85.381 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 427.158 Td /F1 9.8 Tf [(1.00 0.44)] TJ ET 0.267 0.267 0.267 rg 225.054 437.013 194.129 0.750 re f 225.054 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 427.158 Td /F1 9.8 Tf [(0.80 0.24)] TJ ET 0.267 0.267 0.267 rg 418.433 437.013 162.442 0.750 re f 418.433 421.132 0.750 16.631 re f 580.125 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 427.158 Td /F1 9.8 Tf [(20.0%)] TJ ET 0.267 0.267 0.267 rg 26.625 421.132 59.506 0.750 re f 26.625 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 411.277 Td /F1 9.8 Tf [(8)] TJ ET 0.267 0.267 0.267 rg 85.381 421.132 140.423 0.750 re f 85.381 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 411.277 Td /F1 9.8 Tf [(1.06 0.47)] TJ ET 0.267 0.267 0.267 rg 225.054 421.132 194.129 0.750 re f 225.054 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 411.277 Td /F1 9.8 Tf [(0.87 0.27)] TJ ET 0.267 0.267 0.267 rg 418.433 421.132 162.442 0.750 re f 418.433 405.250 0.750 16.631 re f 580.125 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 411.277 Td /F1 9.8 Tf [(17.9%)] TJ ET 0.267 0.267 0.267 rg 26.625 405.250 59.506 0.750 re f 26.625 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 395.395 Td /F1 9.8 Tf [(9)] TJ ET 0.267 0.267 0.267 rg 85.381 405.250 140.423 0.750 re f 85.381 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 395.395 Td /F1 9.8 Tf [(1.07 0.48)] TJ ET 0.267 0.267 0.267 rg 225.054 405.250 194.129 0.750 re f 225.054 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 395.395 Td /F1 9.8 Tf [(0.89 0.28)] TJ ET 0.267 0.267 0.267 rg 418.433 405.250 162.442 0.750 re f 418.433 389.369 0.750 16.631 re f 580.125 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 395.395 Td /F1 9.8 Tf [(16.8%)] TJ ET 0.267 0.267 0.267 rg 26.625 389.369 59.506 0.750 re f 26.625 373.488 59.506 0.750 re f 26.625 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 379.514 Td /F1 9.8 Tf [(10)] TJ ET 0.267 0.267 0.267 rg 85.381 389.369 140.423 0.750 re f 85.381 373.488 140.423 0.750 re f 85.381 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 379.514 Td /F1 9.8 Tf [(1.04 0.44)] TJ ET 0.267 0.267 0.267 rg 225.054 389.369 194.129 0.750 re f 225.054 373.488 194.129 0.750 re f 225.054 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 379.514 Td /F1 9.8 Tf [(0.83 0.30)] TJ ET 0.267 0.267 0.267 rg 418.433 389.369 162.442 0.750 re f 418.433 373.488 162.442 0.750 re f 418.433 373.488 0.750 16.631 re f 580.125 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 379.514 Td /F1 9.8 Tf [(20.2%)] TJ ET BT 26.250 318.589 Td /F1 9.8 Tf [(We next considered the accuracy of forecasting several weeks out \(k > 0\). Table 2 compares predictions based on only )] TJ ET BT 26.250 306.684 Td /F1 9.8 Tf [(historical ILINet data \(the baseline model\) to those enhanced with Twitter data, up to 10 weeks into the future. We found that the )] TJ ET BT 26.250 294.780 Td /F1 9.8 Tf [(Twitter models error after k weeks closely matches the error of the baseline model after k-1 to k-2 weeks. This means that )] TJ ET BT 26.250 282.875 Td /F1 9.8 Tf [(Twitter data provides up to two additional weeks of forecasting ability for a given accuracy tolerance. When attempting to )] TJ ET BT 26.250 270.970 Td /F1 9.8 Tf [(forecast ten weeks into the future \(k=10\), the Twitter model displays less error than the baseline model of four weeks prior. The )] TJ ET BT 26.250 259.065 Td /F1 9.8 Tf [(Twitter model outperforms the baseline for all values of k. In contrast, the baseline model outperforms a model using GFT )] TJ ET BT 26.250 247.161 Td /F1 9.8 Tf [(instead of Twitter for all values of k.)] TJ ET Q q 15.000 169.742 577.500 607.258 re W n q 26.250 763.264 555.000 13.736 re W n 0.271 0.267 0.267 rg BT 26.250 766.011 Td /F1 9.8 Tf [(Table 2. The mean \(+/- SD\) absolute error of two forecasting models after k weeks.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 723.199 555.000 32.565 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 755.764 m 581.250 755.764 l 580.500 755.014 l 27.000 755.014 l f 581.250 755.764 m 581.250 723.199 l 580.500 723.199 l 580.500 755.014 l f 26.250 755.764 m 26.250 723.199 l 27.000 723.199 l 27.000 755.014 l f 0.271 0.267 0.267 rg BT 33.000 741.688 Td /F1 9.0 Tf [(We compared the baseline model based on previous weeks of CDC ILI data to the best-performing model, that incorporating Twitter )] TJ ET BT 33.000 732.530 Td /F1 9.0 Tf [(data. The third column shows the error reduction over the baseline when using the Twitter model.)] TJ ET 1.000 1.000 1.000 rg 26.250 373.113 555.000 350.086 re f 0.965 0.965 0.965 rg 27.000 710.169 58.756 12.280 re f 0.267 0.267 0.267 rg 26.625 722.074 59.131 0.750 re f 26.625 709.794 0.750 13.030 re f 0.271 0.267 0.267 rg BT 31.500 712.925 Td /F4 9.8 Tf [(k)] TJ ET 0.965 0.965 0.965 rg 85.756 710.169 139.673 12.280 re f 0.267 0.267 0.267 rg 85.756 722.074 139.673 0.750 re f 0.271 0.267 0.267 rg BT 90.256 712.925 Td /F4 9.8 Tf [(CDC Only)] TJ ET 0.965 0.965 0.965 rg 225.429 710.169 193.379 12.280 re f 0.267 0.267 0.267 rg 225.429 722.074 193.379 0.750 re f 0.271 0.267 0.267 rg BT 229.929 712.925 Td /F4 9.8 Tf [(CDC+Twitter)] TJ ET 0.965 0.965 0.965 rg 418.808 710.169 161.692 12.280 re f 0.267 0.267 0.267 rg 418.808 722.074 162.067 0.750 re f 580.125 709.794 0.750 13.030 re f 0.271 0.267 0.267 rg BT 423.308 712.925 Td /F4 9.8 Tf [(Reduction)] TJ ET 0.267 0.267 0.267 rg 26.625 709.794 59.506 0.750 re f 26.625 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 699.939 Td /F1 9.8 Tf [(0)] TJ ET 0.267 0.267 0.267 rg 85.381 709.794 140.423 0.750 re f 85.381 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 699.939 Td /F1 9.8 Tf [(0.27 0.06)] TJ ET 0.267 0.267 0.267 rg 225.054 709.794 194.129 0.750 re f 225.054 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 699.939 Td /F1 9.8 Tf [(0.19 0.03)] TJ ET 0.267 0.267 0.267 rg 418.433 709.794 162.442 0.750 re f 418.433 693.913 0.750 16.631 re f 580.125 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 699.939 Td /F1 9.8 Tf [(29.6%)] TJ ET 0.267 0.267 0.267 rg 26.625 693.913 59.506 0.750 re f 26.625 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 684.058 Td /F1 9.8 Tf [(1)] TJ ET 0.267 0.267 0.267 rg 85.381 693.913 140.423 0.750 re f 85.381 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 684.058 Td /F1 9.8 Tf [(0.40 0.12)] TJ ET 0.267 0.267 0.267 rg 225.054 693.913 194.129 0.750 re f 225.054 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 684.058 Td /F1 9.8 Tf [(0.29 0.07)] TJ ET 0.267 0.267 0.267 rg 418.433 693.913 162.442 0.750 re f 418.433 678.031 0.750 16.631 re f 580.125 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 684.058 Td /F1 9.8 Tf [(27.5%)] TJ ET 0.267 0.267 0.267 rg 26.625 678.031 59.506 0.750 re f 26.625 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 668.176 Td /F1 9.8 Tf [(2)] TJ ET 0.267 0.267 0.267 rg 85.381 678.031 140.423 0.750 re f 85.381 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 668.176 Td /F1 9.8 Tf [(0.49 0.17)] TJ ET 0.267 0.267 0.267 rg 225.054 678.031 194.129 0.750 re f 225.054 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 668.176 Td /F1 9.8 Tf [(0.37 0.08)] TJ ET 0.267 0.267 0.267 rg 418.433 678.031 162.442 0.750 re f 418.433 662.150 0.750 16.631 re f 580.125 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 668.176 Td /F1 9.8 Tf [(24.5%)] TJ ET 0.267 0.267 0.267 rg 26.625 662.150 59.506 0.750 re f 26.625 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 652.295 Td /F1 9.8 Tf [(3)] TJ ET 0.267 0.267 0.267 rg 85.381 662.150 140.423 0.750 re f 85.381 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 652.295 Td /F1 9.8 Tf [(0.59 0.22)] TJ ET 0.267 0.267 0.267 rg 225.054 662.150 194.129 0.750 re f 225.054 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 652.295 Td /F1 9.8 Tf [(0.46 0.11)] TJ ET 0.267 0.267 0.267 rg 418.433 662.150 162.442 0.750 re f 418.433 646.269 0.750 16.631 re f 580.125 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 652.295 Td /F1 9.8 Tf [(22.0%)] TJ ET 0.267 0.267 0.267 rg 26.625 646.269 59.506 0.750 re f 26.625 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 636.414 Td /F1 9.8 Tf [(4)] TJ ET 0.267 0.267 0.267 rg 85.381 646.269 140.423 0.750 re f 85.381 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 636.414 Td /F1 9.8 Tf [(0.72 0.27)] TJ ET 0.267 0.267 0.267 rg 225.054 646.269 194.129 0.750 re f 225.054 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 636.414 Td /F1 9.8 Tf [(0.55 0.14)] TJ ET 0.267 0.267 0.267 rg 418.433 646.269 162.442 0.750 re f 418.433 630.388 0.750 16.631 re f 580.125 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 636.414 Td /F1 9.8 Tf [(25.3%)] TJ ET 0.267 0.267 0.267 rg 26.625 630.388 59.506 0.750 re f 26.625 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 31.875 620.533 Td /F1 9.8 Tf [(5)] TJ ET 0.267 0.267 0.267 rg 85.381 630.388 140.423 0.750 re f 85.381 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 90.631 620.533 Td /F1 9.8 Tf [(0.83 0.33)] TJ ET 0.267 0.267 0.267 rg 225.054 630.388 194.129 0.750 re f 225.054 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 230.304 620.533 Td /F1 9.8 Tf [(0.64 0.17)] TJ ET 0.267 0.267 0.267 rg 418.433 630.388 162.442 0.750 re f 418.433 452.894 0.750 178.244 re f 580.125 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 423.683 620.533 Td /F1 9.8 Tf [(23.6%)] TJ ET 0.267 0.267 0.267 rg 26.625 452.894 59.506 0.750 re f 26.625 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 443.039 Td /F1 9.8 Tf [(6)] TJ ET 0.267 0.267 0.267 rg 85.381 452.894 140.423 0.750 re f 85.381 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 443.039 Td /F1 9.8 Tf [(0.92 0.39)] TJ ET 0.267 0.267 0.267 rg 225.054 452.894 194.129 0.750 re f 225.054 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 443.039 Td /F1 9.8 Tf [(0.73 0.21)] TJ ET 0.267 0.267 0.267 rg 418.433 452.894 162.442 0.750 re f 418.433 437.013 0.750 16.631 re f 580.125 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 443.039 Td /F1 9.8 Tf [(20.7%)] TJ ET 0.267 0.267 0.267 rg 26.625 437.013 59.506 0.750 re f 26.625 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 427.158 Td /F1 9.8 Tf [(7)] TJ ET 0.267 0.267 0.267 rg 85.381 437.013 140.423 0.750 re f 85.381 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 427.158 Td /F1 9.8 Tf [(1.00 0.44)] TJ ET 0.267 0.267 0.267 rg 225.054 437.013 194.129 0.750 re f 225.054 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 427.158 Td /F1 9.8 Tf [(0.80 0.24)] TJ ET 0.267 0.267 0.267 rg 418.433 437.013 162.442 0.750 re f 418.433 421.132 0.750 16.631 re f 580.125 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 427.158 Td /F1 9.8 Tf [(20.0%)] TJ ET 0.267 0.267 0.267 rg 26.625 421.132 59.506 0.750 re f 26.625 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 411.277 Td /F1 9.8 Tf [(8)] TJ ET 0.267 0.267 0.267 rg 85.381 421.132 140.423 0.750 re f 85.381 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 411.277 Td /F1 9.8 Tf [(1.06 0.47)] TJ ET 0.267 0.267 0.267 rg 225.054 421.132 194.129 0.750 re f 225.054 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 411.277 Td /F1 9.8 Tf [(0.87 0.27)] TJ ET 0.267 0.267 0.267 rg 418.433 421.132 162.442 0.750 re f 418.433 405.250 0.750 16.631 re f 580.125 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 411.277 Td /F1 9.8 Tf [(17.9%)] TJ ET 0.267 0.267 0.267 rg 26.625 405.250 59.506 0.750 re f 26.625 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 395.395 Td /F1 9.8 Tf [(9)] TJ ET 0.267 0.267 0.267 rg 85.381 405.250 140.423 0.750 re f 85.381 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 395.395 Td /F1 9.8 Tf [(1.07 0.48)] TJ ET 0.267 0.267 0.267 rg 225.054 405.250 194.129 0.750 re f 225.054 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 395.395 Td /F1 9.8 Tf [(0.89 0.28)] TJ ET 0.267 0.267 0.267 rg 418.433 405.250 162.442 0.750 re f 418.433 389.369 0.750 16.631 re f 580.125 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 395.395 Td /F1 9.8 Tf [(16.8%)] TJ ET 0.267 0.267 0.267 rg 26.625 389.369 59.506 0.750 re f 26.625 373.488 59.506 0.750 re f 26.625 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 379.514 Td /F1 9.8 Tf [(10)] TJ ET 0.267 0.267 0.267 rg 85.381 389.369 140.423 0.750 re f 85.381 373.488 140.423 0.750 re f 85.381 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 379.514 Td /F1 9.8 Tf [(1.04 0.44)] TJ ET 0.267 0.267 0.267 rg 225.054 389.369 194.129 0.750 re f 225.054 373.488 194.129 0.750 re f 225.054 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 379.514 Td /F1 9.8 Tf [(0.83 0.30)] TJ ET 0.267 0.267 0.267 rg 418.433 389.369 162.442 0.750 re f 418.433 373.488 162.442 0.750 re f 418.433 373.488 0.750 16.631 re f 580.125 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 379.514 Td /F1 9.8 Tf [(20.2%)] TJ ET BT 26.250 318.589 Td /F1 9.8 Tf [(We next considered the accuracy of forecasting several weeks out \(k > 0\). Table 2 compares predictions based on only )] TJ ET BT 26.250 306.684 Td /F1 9.8 Tf [(historical ILINet data \(the baseline model\) to those enhanced with Twitter data, up to 10 weeks into the future. We found that the )] TJ ET BT 26.250 294.780 Td /F1 9.8 Tf [(Twitter models error after k weeks closely matches the error of the baseline model after k-1 to k-2 weeks. This means that )] TJ ET BT 26.250 282.875 Td /F1 9.8 Tf [(Twitter data provides up to two additional weeks of forecasting ability for a given accuracy tolerance. When attempting to )] TJ ET BT 26.250 270.970 Td /F1 9.8 Tf [(forecast ten weeks into the future \(k=10\), the Twitter model displays less error than the baseline model of four weeks prior. The )] TJ ET BT 26.250 259.065 Td /F1 9.8 Tf [(Twitter model outperforms the baseline for all values of k. In contrast, the baseline model outperforms a model using GFT )] TJ ET BT 26.250 247.161 Td /F1 9.8 Tf [(instead of Twitter for all values of k.)] TJ ET Q q 15.000 169.742 577.500 607.258 re W n q 26.250 763.264 555.000 13.736 re W n 0.271 0.267 0.267 rg BT 26.250 766.011 Td /F1 9.8 Tf [(Table 2. The mean \(+/- SD\) absolute error of two forecasting models after k weeks.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 723.199 555.000 32.565 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 755.764 m 581.250 755.764 l 580.500 755.014 l 27.000 755.014 l f 581.250 755.764 m 581.250 723.199 l 580.500 723.199 l 580.500 755.014 l f 26.250 755.764 m 26.250 723.199 l 27.000 723.199 l 27.000 755.014 l f 0.271 0.267 0.267 rg BT 33.000 741.688 Td /F1 9.0 Tf [(We compared the baseline model based on previous weeks of CDC ILI data to the best-performing model, that incorporating Twitter )] TJ ET BT 33.000 732.530 Td /F1 9.0 Tf [(data. The third column shows the error reduction over the baseline when using the Twitter model.)] TJ ET 1.000 1.000 1.000 rg 26.250 373.113 555.000 350.086 re f 0.965 0.965 0.965 rg 27.000 710.169 58.756 12.280 re f 0.267 0.267 0.267 rg 26.625 722.074 59.131 0.750 re f 26.625 709.794 0.750 13.030 re f 0.271 0.267 0.267 rg BT 31.500 712.925 Td /F4 9.8 Tf [(k)] TJ ET 0.965 0.965 0.965 rg 85.756 710.169 139.673 12.280 re f 0.267 0.267 0.267 rg 85.756 722.074 139.673 0.750 re f 0.271 0.267 0.267 rg BT 90.256 712.925 Td /F4 9.8 Tf [(CDC Only)] TJ ET 0.965 0.965 0.965 rg 225.429 710.169 193.379 12.280 re f 0.267 0.267 0.267 rg 225.429 722.074 193.379 0.750 re f 0.271 0.267 0.267 rg BT 229.929 712.925 Td /F4 9.8 Tf [(CDC+Twitter)] TJ ET 0.965 0.965 0.965 rg 418.808 710.169 161.692 12.280 re f 0.267 0.267 0.267 rg 418.808 722.074 162.067 0.750 re f 580.125 709.794 0.750 13.030 re f 0.271 0.267 0.267 rg BT 423.308 712.925 Td /F4 9.8 Tf [(Reduction)] TJ ET 0.267 0.267 0.267 rg 26.625 709.794 59.506 0.750 re f 26.625 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 699.939 Td /F1 9.8 Tf [(0)] TJ ET 0.267 0.267 0.267 rg 85.381 709.794 140.423 0.750 re f 85.381 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 699.939 Td /F1 9.8 Tf [(0.27 0.06)] TJ ET 0.267 0.267 0.267 rg 225.054 709.794 194.129 0.750 re f 225.054 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 699.939 Td /F1 9.8 Tf [(0.19 0.03)] TJ ET 0.267 0.267 0.267 rg 418.433 709.794 162.442 0.750 re f 418.433 693.913 0.750 16.631 re f 580.125 693.913 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 699.939 Td /F1 9.8 Tf [(29.6%)] TJ ET 0.267 0.267 0.267 rg 26.625 693.913 59.506 0.750 re f 26.625 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 684.058 Td /F1 9.8 Tf [(1)] TJ ET 0.267 0.267 0.267 rg 85.381 693.913 140.423 0.750 re f 85.381 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 684.058 Td /F1 9.8 Tf [(0.40 0.12)] TJ ET 0.267 0.267 0.267 rg 225.054 693.913 194.129 0.750 re f 225.054 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 684.058 Td /F1 9.8 Tf [(0.29 0.07)] TJ ET 0.267 0.267 0.267 rg 418.433 693.913 162.442 0.750 re f 418.433 678.031 0.750 16.631 re f 580.125 678.031 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 684.058 Td /F1 9.8 Tf [(27.5%)] TJ ET 0.267 0.267 0.267 rg 26.625 678.031 59.506 0.750 re f 26.625 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 668.176 Td /F1 9.8 Tf [(2)] TJ ET 0.267 0.267 0.267 rg 85.381 678.031 140.423 0.750 re f 85.381 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 668.176 Td /F1 9.8 Tf [(0.49 0.17)] TJ ET 0.267 0.267 0.267 rg 225.054 678.031 194.129 0.750 re f 225.054 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 668.176 Td /F1 9.8 Tf [(0.37 0.08)] TJ ET 0.267 0.267 0.267 rg 418.433 678.031 162.442 0.750 re f 418.433 662.150 0.750 16.631 re f 580.125 662.150 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 668.176 Td /F1 9.8 Tf [(24.5%)] TJ ET 0.267 0.267 0.267 rg 26.625 662.150 59.506 0.750 re f 26.625 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 652.295 Td /F1 9.8 Tf [(3)] TJ ET 0.267 0.267 0.267 rg 85.381 662.150 140.423 0.750 re f 85.381 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 652.295 Td /F1 9.8 Tf [(0.59 0.22)] TJ ET 0.267 0.267 0.267 rg 225.054 662.150 194.129 0.750 re f 225.054 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 652.295 Td /F1 9.8 Tf [(0.46 0.11)] TJ ET 0.267 0.267 0.267 rg 418.433 662.150 162.442 0.750 re f 418.433 646.269 0.750 16.631 re f 580.125 646.269 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 652.295 Td /F1 9.8 Tf [(22.0%)] TJ ET 0.267 0.267 0.267 rg 26.625 646.269 59.506 0.750 re f 26.625 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 636.414 Td /F1 9.8 Tf [(4)] TJ ET 0.267 0.267 0.267 rg 85.381 646.269 140.423 0.750 re f 85.381 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 636.414 Td /F1 9.8 Tf [(0.72 0.27)] TJ ET 0.267 0.267 0.267 rg 225.054 646.269 194.129 0.750 re f 225.054 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 636.414 Td /F1 9.8 Tf [(0.55 0.14)] TJ ET 0.267 0.267 0.267 rg 418.433 646.269 162.442 0.750 re f 418.433 630.388 0.750 16.631 re f 580.125 630.388 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 636.414 Td /F1 9.8 Tf [(25.3%)] TJ ET 0.267 0.267 0.267 rg 26.625 630.388 59.506 0.750 re f 26.625 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 31.875 620.533 Td /F1 9.8 Tf [(5)] TJ ET 0.267 0.267 0.267 rg 85.381 630.388 140.423 0.750 re f 85.381 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 90.631 620.533 Td /F1 9.8 Tf [(0.83 0.33)] TJ ET 0.267 0.267 0.267 rg 225.054 630.388 194.129 0.750 re f 225.054 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 230.304 620.533 Td /F1 9.8 Tf [(0.64 0.17)] TJ ET 0.267 0.267 0.267 rg 418.433 630.388 162.442 0.750 re f 418.433 452.894 0.750 178.244 re f 580.125 452.894 0.750 178.244 re f 0.271 0.267 0.267 rg BT 423.683 620.533 Td /F1 9.8 Tf [(23.6%)] TJ ET 0.267 0.267 0.267 rg 26.625 452.894 59.506 0.750 re f 26.625 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 443.039 Td /F1 9.8 Tf [(6)] TJ ET 0.267 0.267 0.267 rg 85.381 452.894 140.423 0.750 re f 85.381 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 443.039 Td /F1 9.8 Tf [(0.92 0.39)] TJ ET 0.267 0.267 0.267 rg 225.054 452.894 194.129 0.750 re f 225.054 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 443.039 Td /F1 9.8 Tf [(0.73 0.21)] TJ ET 0.267 0.267 0.267 rg 418.433 452.894 162.442 0.750 re f 418.433 437.013 0.750 16.631 re f 580.125 437.013 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 443.039 Td /F1 9.8 Tf [(20.7%)] TJ ET 0.267 0.267 0.267 rg 26.625 437.013 59.506 0.750 re f 26.625 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 427.158 Td /F1 9.8 Tf [(7)] TJ ET 0.267 0.267 0.267 rg 85.381 437.013 140.423 0.750 re f 85.381 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 427.158 Td /F1 9.8 Tf [(1.00 0.44)] TJ ET 0.267 0.267 0.267 rg 225.054 437.013 194.129 0.750 re f 225.054 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 427.158 Td /F1 9.8 Tf [(0.80 0.24)] TJ ET 0.267 0.267 0.267 rg 418.433 437.013 162.442 0.750 re f 418.433 421.132 0.750 16.631 re f 580.125 421.132 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 427.158 Td /F1 9.8 Tf [(20.0%)] TJ ET 0.267 0.267 0.267 rg 26.625 421.132 59.506 0.750 re f 26.625 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 411.277 Td /F1 9.8 Tf [(8)] TJ ET 0.267 0.267 0.267 rg 85.381 421.132 140.423 0.750 re f 85.381 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 411.277 Td /F1 9.8 Tf [(1.06 0.47)] TJ ET 0.267 0.267 0.267 rg 225.054 421.132 194.129 0.750 re f 225.054 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 411.277 Td /F1 9.8 Tf [(0.87 0.27)] TJ ET 0.267 0.267 0.267 rg 418.433 421.132 162.442 0.750 re f 418.433 405.250 0.750 16.631 re f 580.125 405.250 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 411.277 Td /F1 9.8 Tf [(17.9%)] TJ ET 0.267 0.267 0.267 rg 26.625 405.250 59.506 0.750 re f 26.625 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 395.395 Td /F1 9.8 Tf [(9)] TJ ET 0.267 0.267 0.267 rg 85.381 405.250 140.423 0.750 re f 85.381 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 395.395 Td /F1 9.8 Tf [(1.07 0.48)] TJ ET 0.267 0.267 0.267 rg 225.054 405.250 194.129 0.750 re f 225.054 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 395.395 Td /F1 9.8 Tf [(0.89 0.28)] TJ ET 0.267 0.267 0.267 rg 418.433 405.250 162.442 0.750 re f 418.433 389.369 0.750 16.631 re f 580.125 389.369 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 395.395 Td /F1 9.8 Tf [(16.8%)] TJ ET 0.267 0.267 0.267 rg 26.625 389.369 59.506 0.750 re f 26.625 373.488 59.506 0.750 re f 26.625 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 379.514 Td /F1 9.8 Tf [(10)] TJ ET 0.267 0.267 0.267 rg 85.381 389.369 140.423 0.750 re f 85.381 373.488 140.423 0.750 re f 85.381 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 90.631 379.514 Td /F1 9.8 Tf [(1.04 0.44)] TJ ET 0.267 0.267 0.267 rg 225.054 389.369 194.129 0.750 re f 225.054 373.488 194.129 0.750 re f 225.054 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 230.304 379.514 Td /F1 9.8 Tf [(0.83 0.30)] TJ ET 0.267 0.267 0.267 rg 418.433 389.369 162.442 0.750 re f 418.433 373.488 162.442 0.750 re f 418.433 373.488 0.750 16.631 re f 580.125 373.488 0.750 16.631 re f 0.271 0.267 0.267 rg BT 423.683 379.514 Td /F1 9.8 Tf [(20.2%)] TJ ET BT 26.250 318.589 Td /F1 9.8 Tf [(We next considered the accuracy of forecasting several weeks out \(k > 0\). Table 2 compares predictions based on only )] TJ ET BT 26.250 306.684 Td /F1 9.8 Tf [(historical ILINet data \(the baseline model\) to those enhanced with Twitter data, up to 10 weeks into the future. We found that the )] TJ ET BT 26.250 294.780 Td /F1 9.8 Tf [(Twitter models error after k weeks closely matches the error of the baseline model after k-1 to k-2 weeks. This means that )] TJ ET BT 26.250 282.875 Td /F1 9.8 Tf [(Twitter data provides up to two additional weeks of forecasting ability for a given accuracy tolerance. When attempting to )] TJ ET BT 26.250 270.970 Td /F1 9.8 Tf [(forecast ten weeks into the future \(k=10\), the Twitter model displays less error than the baseline model of four weeks prior. The )] TJ ET BT 26.250 259.065 Td /F1 9.8 Tf [(Twitter model outperforms the baseline for all values of k. In contrast, the baseline model outperforms a model using GFT )] TJ ET BT 26.250 247.161 Td /F1 9.8 Tf [(instead of Twitter for all values of k.)] TJ ET Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(4)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj 357 0 obj << /Type /Page /Parent 3 0 R /Annots [ 359 0 R 363 0 R 367 0 R ] /Contents 358 0 R >> endobj 358 0 obj << /Length 26232 >> stream 0.271 0.267 0.267 rg 0.267 0.267 0.267 RG q 15.000 82.642 577.500 694.358 re W n q 26.250 749.527 555.000 27.473 re W n 0.271 0.267 0.267 rg BT 26.250 766.011 Td /F1 9.8 Tf [(Table 3. Summary of revisions made to CDC ILINet data after k weeks, where k=0 corresponds to the first value reported for a )] TJ ET BT 26.250 752.275 Td /F1 9.8 Tf [(given week.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 709.462 555.000 32.565 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 742.027 m 581.250 742.027 l 580.500 741.277 l 27.000 741.277 l f 581.250 742.027 m 581.250 709.462 l 580.500 709.462 l 580.500 741.277 l f 26.250 742.027 m 26.250 709.462 l 27.000 709.462 l 27.000 741.277 l f 0.271 0.267 0.267 rg BT 33.000 727.951 Td /F1 9.0 Tf [(We measured the mean absolute difference \(MAD\) and mean difference \(MD\) between the original forecast and the revision after k )] TJ ET BT 33.000 718.794 Td /F1 9.0 Tf [(weeks. The difference from the weeks previous values after k-1 weeks is also shown.)] TJ ET 1.000 1.000 1.000 rg 26.250 409.165 555.000 300.297 re f 0.965 0.965 0.965 rg 27.000 696.433 30.500 12.280 re f 0.267 0.267 0.267 rg 26.625 708.337 30.875 0.750 re f 26.625 696.058 0.750 13.030 re f 0.965 0.965 0.965 rg 57.500 696.433 177.551 12.280 re f 0.267 0.267 0.267 rg 57.500 708.337 177.551 0.750 re f 0.271 0.267 0.267 rg BT 62.000 699.189 Td /F4 9.8 Tf [(Change From Final Report)] TJ ET 0.965 0.965 0.965 rg 235.050 696.433 71.925 12.280 re f 0.267 0.267 0.267 rg 235.050 708.337 71.925 0.750 re f 0.965 0.965 0.965 rg 306.975 696.433 201.601 12.280 re f 0.267 0.267 0.267 rg 306.975 708.337 201.601 0.750 re f 0.271 0.267 0.267 rg BT 311.475 699.189 Td /F4 9.8 Tf [(Change From Previous Report)] TJ ET 0.965 0.965 0.965 rg 508.575 696.433 71.925 12.280 re f 0.267 0.267 0.267 rg 508.575 708.337 72.300 0.750 re f 580.125 696.058 0.750 13.030 re f 26.625 696.058 31.250 0.750 re f 26.625 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 686.203 Td /F4 9.8 Tf [(k)] TJ ET 0.267 0.267 0.267 rg 57.125 696.058 178.301 0.750 re f 57.125 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 686.203 Td /F4 9.8 Tf [(MAD)] TJ ET 0.267 0.267 0.267 rg 234.675 696.058 72.675 0.750 re f 234.675 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 686.203 Td /F4 9.8 Tf [(MD)] TJ ET 0.267 0.267 0.267 rg 306.600 696.058 202.351 0.750 re f 306.600 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 686.203 Td /F4 9.8 Tf [(MAD)] TJ ET 0.267 0.267 0.267 rg 508.200 696.058 72.675 0.750 re f 508.200 680.176 0.750 16.631 re f 580.125 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 686.203 Td /F4 9.8 Tf [(MD)] TJ ET 0.267 0.267 0.267 rg 26.625 680.176 31.250 0.750 re f 26.625 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 670.321 Td /F1 9.8 Tf [(0)] TJ ET 0.267 0.267 0.267 rg 57.125 680.176 178.301 0.750 re f 57.125 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 670.321 Td /F1 9.8 Tf [(0.137)] TJ ET 0.267 0.267 0.267 rg 234.675 680.176 72.675 0.750 re f 234.675 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 670.321 Td /F1 9.8 Tf [(-0.030)] TJ ET 0.267 0.267 0.267 rg 306.600 680.176 202.351 0.750 re f 306.600 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 670.321 Td /F1 9.8 Tf [(n/a)] TJ ET 0.267 0.267 0.267 rg 508.200 680.176 72.675 0.750 re f 508.200 664.295 0.750 16.631 re f 580.125 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 670.321 Td /F1 9.8 Tf [(n/a)] TJ ET 0.267 0.267 0.267 rg 26.625 664.295 31.250 0.750 re f 26.625 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 654.440 Td /F1 9.8 Tf [(1)] TJ ET 0.267 0.267 0.267 rg 57.125 664.295 178.301 0.750 re f 57.125 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 654.440 Td /F1 9.8 Tf [(0.083)] TJ ET 0.267 0.267 0.267 rg 234.675 664.295 72.675 0.750 re f 234.675 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 654.440 Td /F1 9.8 Tf [(-0.002)] TJ ET 0.267 0.267 0.267 rg 306.600 664.295 202.351 0.750 re f 306.600 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 654.440 Td /F1 9.8 Tf [(0.101)] TJ ET 0.267 0.267 0.267 rg 508.200 664.295 72.675 0.750 re f 508.200 648.414 0.750 16.631 re f 580.125 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 654.440 Td /F1 9.8 Tf [(0.028)] TJ ET 0.267 0.267 0.267 rg 26.625 648.414 31.250 0.750 re f 26.625 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 638.559 Td /F1 9.8 Tf [(2)] TJ ET 0.267 0.267 0.267 rg 57.125 648.414 178.301 0.750 re f 57.125 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 638.559 Td /F1 9.8 Tf [(0.082)] TJ ET 0.267 0.267 0.267 rg 234.675 648.414 72.675 0.750 re f 234.675 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 638.559 Td /F1 9.8 Tf [(-0.008)] TJ ET 0.267 0.267 0.267 rg 306.600 648.414 202.351 0.750 re f 306.600 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 638.559 Td /F1 9.8 Tf [(0.073)] TJ ET 0.267 0.267 0.267 rg 508.200 648.414 72.675 0.750 re f 508.200 632.533 0.750 16.631 re f 580.125 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 638.559 Td /F1 9.8 Tf [(-0.007)] TJ ET 0.267 0.267 0.267 rg 26.625 632.533 31.250 0.750 re f 26.625 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 622.678 Td /F1 9.8 Tf [(3)] TJ ET 0.267 0.267 0.267 rg 57.125 632.533 178.301 0.750 re f 57.125 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 622.678 Td /F1 9.8 Tf [(0.060)] TJ ET 0.267 0.267 0.267 rg 234.675 632.533 72.675 0.750 re f 234.675 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 622.678 Td /F1 9.8 Tf [(0.003)] TJ ET 0.267 0.267 0.267 rg 306.600 632.533 202.351 0.750 re f 306.600 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 622.678 Td /F1 9.8 Tf [(0.050)] TJ ET 0.267 0.267 0.267 rg 508.200 632.533 72.675 0.750 re f 508.200 616.651 0.750 16.631 re f 580.125 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 622.678 Td /F1 9.8 Tf [(0.011)] TJ ET 0.267 0.267 0.267 rg 26.625 616.651 31.250 0.750 re f 26.625 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 31.875 606.796 Td /F1 9.8 Tf [(4)] TJ ET 0.267 0.267 0.267 rg 57.125 616.651 178.301 0.750 re f 57.125 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 62.375 606.796 Td /F1 9.8 Tf [(0.053)] TJ ET 0.267 0.267 0.267 rg 234.675 616.651 72.675 0.750 re f 234.675 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 239.925 606.796 Td /F1 9.8 Tf [(-0.001)] TJ ET 0.267 0.267 0.267 rg 306.600 616.651 202.351 0.750 re f 306.600 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 311.850 606.796 Td /F1 9.8 Tf [(0.027)] TJ ET 0.267 0.267 0.267 rg 508.200 616.651 72.675 0.750 re f 508.200 425.422 0.750 191.980 re f 580.125 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 513.450 606.796 Td /F1 9.8 Tf [(-0.004)] TJ ET 0.267 0.267 0.267 rg 26.625 425.422 31.250 0.750 re f 26.625 409.540 31.250 0.750 re f 26.625 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 415.567 Td /F1 9.8 Tf [(5)] TJ ET 0.267 0.267 0.267 rg 57.125 425.422 178.301 0.750 re f 57.125 409.540 178.301 0.750 re f 57.125 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 415.567 Td /F1 9.8 Tf [(0.048)] TJ ET 0.267 0.267 0.267 rg 234.675 425.422 72.675 0.750 re f 234.675 409.540 72.675 0.750 re f 234.675 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 415.567 Td /F1 9.8 Tf [(-0.004)] TJ ET 0.267 0.267 0.267 rg 306.600 425.422 202.351 0.750 re f 306.600 409.540 202.351 0.750 re f 306.600 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 415.567 Td /F1 9.8 Tf [(0.020)] TJ ET 0.267 0.267 0.267 rg 508.200 425.422 72.675 0.750 re f 508.200 409.540 72.675 0.750 re f 508.200 409.540 0.750 16.631 re f 580.125 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 415.567 Td /F1 9.8 Tf [(-0.003)] TJ ET BT 26.250 354.642 Td /F1 9.8 Tf [(Finally, we measured the extent to which ILINet revisions impact the efficacy of forecasts. In addition, we measured the )] TJ ET BT 26.250 342.737 Td /F1 9.8 Tf [(differences in a given weeks ILINet values for subsequent reports \(Table 3\). We found that the magnitude of these differences )] TJ ET BT 26.250 330.832 Td /F1 9.8 Tf [(is, on average, 12% of the standard deviation of the final values, with the average error gradually decreasing as more data )] TJ ET BT 26.250 318.927 Td /F1 9.8 Tf [(become available over time. Nevertheless, the most recently available ILINet data values are, on average, inaccurate.)] TJ ET 0.965 0.965 0.965 rg 26.250 82.642 555.000 226.405 re f 0.267 0.267 0.267 rg 26.250 309.046 m 581.250 309.046 l 581.250 308.296 l 26.250 308.296 l f q 450.000 0 0 192.750 35.250 106.546 cm /I46 Do Q q 35.250 82.642 537.000 17.905 re W n Q Q q 15.000 82.642 577.500 694.358 re W n q 26.250 749.527 555.000 27.473 re W n 0.271 0.267 0.267 rg BT 26.250 766.011 Td /F1 9.8 Tf [(Table 3. Summary of revisions made to CDC ILINet data after k weeks, where k=0 corresponds to the first value reported for a )] TJ ET BT 26.250 752.275 Td /F1 9.8 Tf [(given week.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 709.462 555.000 32.565 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 742.027 m 581.250 742.027 l 580.500 741.277 l 27.000 741.277 l f 581.250 742.027 m 581.250 709.462 l 580.500 709.462 l 580.500 741.277 l f 26.250 742.027 m 26.250 709.462 l 27.000 709.462 l 27.000 741.277 l f 0.271 0.267 0.267 rg BT 33.000 727.951 Td /F1 9.0 Tf [(We measured the mean absolute difference \(MAD\) and mean difference \(MD\) between the original forecast and the revision after k )] TJ ET BT 33.000 718.794 Td /F1 9.0 Tf [(weeks. The difference from the weeks previous values after k-1 weeks is also shown.)] TJ ET 1.000 1.000 1.000 rg 26.250 409.165 555.000 300.297 re f 0.965 0.965 0.965 rg 27.000 696.433 30.500 12.280 re f 0.267 0.267 0.267 rg 26.625 708.337 30.875 0.750 re f 26.625 696.058 0.750 13.030 re f 0.965 0.965 0.965 rg 57.500 696.433 177.551 12.280 re f 0.267 0.267 0.267 rg 57.500 708.337 177.551 0.750 re f 0.271 0.267 0.267 rg BT 62.000 699.189 Td /F4 9.8 Tf [(Change From Final Report)] TJ ET 0.965 0.965 0.965 rg 235.050 696.433 71.925 12.280 re f 0.267 0.267 0.267 rg 235.050 708.337 71.925 0.750 re f 0.965 0.965 0.965 rg 306.975 696.433 201.601 12.280 re f 0.267 0.267 0.267 rg 306.975 708.337 201.601 0.750 re f 0.271 0.267 0.267 rg BT 311.475 699.189 Td /F4 9.8 Tf [(Change From Previous Report)] TJ ET 0.965 0.965 0.965 rg 508.575 696.433 71.925 12.280 re f 0.267 0.267 0.267 rg 508.575 708.337 72.300 0.750 re f 580.125 696.058 0.750 13.030 re f 26.625 696.058 31.250 0.750 re f 26.625 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 686.203 Td /F4 9.8 Tf [(k)] TJ ET 0.267 0.267 0.267 rg 57.125 696.058 178.301 0.750 re f 57.125 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 686.203 Td /F4 9.8 Tf [(MAD)] TJ ET 0.267 0.267 0.267 rg 234.675 696.058 72.675 0.750 re f 234.675 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 686.203 Td /F4 9.8 Tf [(MD)] TJ ET 0.267 0.267 0.267 rg 306.600 696.058 202.351 0.750 re f 306.600 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 686.203 Td /F4 9.8 Tf [(MAD)] TJ ET 0.267 0.267 0.267 rg 508.200 696.058 72.675 0.750 re f 508.200 680.176 0.750 16.631 re f 580.125 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 686.203 Td /F4 9.8 Tf [(MD)] TJ ET 0.267 0.267 0.267 rg 26.625 680.176 31.250 0.750 re f 26.625 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 670.321 Td /F1 9.8 Tf [(0)] TJ ET 0.267 0.267 0.267 rg 57.125 680.176 178.301 0.750 re f 57.125 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 670.321 Td /F1 9.8 Tf [(0.137)] TJ ET 0.267 0.267 0.267 rg 234.675 680.176 72.675 0.750 re f 234.675 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 670.321 Td /F1 9.8 Tf [(-0.030)] TJ ET 0.267 0.267 0.267 rg 306.600 680.176 202.351 0.750 re f 306.600 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 670.321 Td /F1 9.8 Tf [(n/a)] TJ ET 0.267 0.267 0.267 rg 508.200 680.176 72.675 0.750 re f 508.200 664.295 0.750 16.631 re f 580.125 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 670.321 Td /F1 9.8 Tf [(n/a)] TJ ET 0.267 0.267 0.267 rg 26.625 664.295 31.250 0.750 re f 26.625 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 654.440 Td /F1 9.8 Tf [(1)] TJ ET 0.267 0.267 0.267 rg 57.125 664.295 178.301 0.750 re f 57.125 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 654.440 Td /F1 9.8 Tf [(0.083)] TJ ET 0.267 0.267 0.267 rg 234.675 664.295 72.675 0.750 re f 234.675 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 654.440 Td /F1 9.8 Tf [(-0.002)] TJ ET 0.267 0.267 0.267 rg 306.600 664.295 202.351 0.750 re f 306.600 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 654.440 Td /F1 9.8 Tf [(0.101)] TJ ET 0.267 0.267 0.267 rg 508.200 664.295 72.675 0.750 re f 508.200 648.414 0.750 16.631 re f 580.125 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 654.440 Td /F1 9.8 Tf [(0.028)] TJ ET 0.267 0.267 0.267 rg 26.625 648.414 31.250 0.750 re f 26.625 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 638.559 Td /F1 9.8 Tf [(2)] TJ ET 0.267 0.267 0.267 rg 57.125 648.414 178.301 0.750 re f 57.125 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 638.559 Td /F1 9.8 Tf [(0.082)] TJ ET 0.267 0.267 0.267 rg 234.675 648.414 72.675 0.750 re f 234.675 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 638.559 Td /F1 9.8 Tf [(-0.008)] TJ ET 0.267 0.267 0.267 rg 306.600 648.414 202.351 0.750 re f 306.600 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 638.559 Td /F1 9.8 Tf [(0.073)] TJ ET 0.267 0.267 0.267 rg 508.200 648.414 72.675 0.750 re f 508.200 632.533 0.750 16.631 re f 580.125 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 638.559 Td /F1 9.8 Tf [(-0.007)] TJ ET 0.267 0.267 0.267 rg 26.625 632.533 31.250 0.750 re f 26.625 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 622.678 Td /F1 9.8 Tf [(3)] TJ ET 0.267 0.267 0.267 rg 57.125 632.533 178.301 0.750 re f 57.125 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 622.678 Td /F1 9.8 Tf [(0.060)] TJ ET 0.267 0.267 0.267 rg 234.675 632.533 72.675 0.750 re f 234.675 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 622.678 Td /F1 9.8 Tf [(0.003)] TJ ET 0.267 0.267 0.267 rg 306.600 632.533 202.351 0.750 re f 306.600 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 622.678 Td /F1 9.8 Tf [(0.050)] TJ ET 0.267 0.267 0.267 rg 508.200 632.533 72.675 0.750 re f 508.200 616.651 0.750 16.631 re f 580.125 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 622.678 Td /F1 9.8 Tf [(0.011)] TJ ET 0.267 0.267 0.267 rg 26.625 616.651 31.250 0.750 re f 26.625 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 31.875 606.796 Td /F1 9.8 Tf [(4)] TJ ET 0.267 0.267 0.267 rg 57.125 616.651 178.301 0.750 re f 57.125 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 62.375 606.796 Td /F1 9.8 Tf [(0.053)] TJ ET 0.267 0.267 0.267 rg 234.675 616.651 72.675 0.750 re f 234.675 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 239.925 606.796 Td /F1 9.8 Tf [(-0.001)] TJ ET 0.267 0.267 0.267 rg 306.600 616.651 202.351 0.750 re f 306.600 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 311.850 606.796 Td /F1 9.8 Tf [(0.027)] TJ ET 0.267 0.267 0.267 rg 508.200 616.651 72.675 0.750 re f 508.200 425.422 0.750 191.980 re f 580.125 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 513.450 606.796 Td /F1 9.8 Tf [(-0.004)] TJ ET 0.267 0.267 0.267 rg 26.625 425.422 31.250 0.750 re f 26.625 409.540 31.250 0.750 re f 26.625 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 415.567 Td /F1 9.8 Tf [(5)] TJ ET 0.267 0.267 0.267 rg 57.125 425.422 178.301 0.750 re f 57.125 409.540 178.301 0.750 re f 57.125 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 415.567 Td /F1 9.8 Tf [(0.048)] TJ ET 0.267 0.267 0.267 rg 234.675 425.422 72.675 0.750 re f 234.675 409.540 72.675 0.750 re f 234.675 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 415.567 Td /F1 9.8 Tf [(-0.004)] TJ ET 0.267 0.267 0.267 rg 306.600 425.422 202.351 0.750 re f 306.600 409.540 202.351 0.750 re f 306.600 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 415.567 Td /F1 9.8 Tf [(0.020)] TJ ET 0.267 0.267 0.267 rg 508.200 425.422 72.675 0.750 re f 508.200 409.540 72.675 0.750 re f 508.200 409.540 0.750 16.631 re f 580.125 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 415.567 Td /F1 9.8 Tf [(-0.003)] TJ ET BT 26.250 354.642 Td /F1 9.8 Tf [(Finally, we measured the extent to which ILINet revisions impact the efficacy of forecasts. In addition, we measured the )] TJ ET BT 26.250 342.737 Td /F1 9.8 Tf [(differences in a given weeks ILINet values for subsequent reports \(Table 3\). We found that the magnitude of these differences )] TJ ET BT 26.250 330.832 Td /F1 9.8 Tf [(is, on average, 12% of the standard deviation of the final values, with the average error gradually decreasing as more data )] TJ ET BT 26.250 318.927 Td /F1 9.8 Tf [(become available over time. Nevertheless, the most recently available ILINet data values are, on average, inaccurate.)] TJ ET 0.965 0.965 0.965 rg 26.250 82.642 555.000 226.405 re f 0.267 0.267 0.267 rg 26.250 309.046 m 581.250 309.046 l 581.250 308.296 l 26.250 308.296 l f q 450.000 0 0 192.750 35.250 106.546 cm /I48 Do Q q 35.250 82.642 537.000 17.905 re W n Q Q q 15.000 82.642 577.500 694.358 re W n q 26.250 749.527 555.000 27.473 re W n 0.271 0.267 0.267 rg BT 26.250 766.011 Td /F1 9.8 Tf [(Table 3. Summary of revisions made to CDC ILINet data after k weeks, where k=0 corresponds to the first value reported for a )] TJ ET BT 26.250 752.275 Td /F1 9.8 Tf [(given week.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 709.462 555.000 32.565 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 742.027 m 581.250 742.027 l 580.500 741.277 l 27.000 741.277 l f 581.250 742.027 m 581.250 709.462 l 580.500 709.462 l 580.500 741.277 l f 26.250 742.027 m 26.250 709.462 l 27.000 709.462 l 27.000 741.277 l f 0.271 0.267 0.267 rg BT 33.000 727.951 Td /F1 9.0 Tf [(We measured the mean absolute difference \(MAD\) and mean difference \(MD\) between the original forecast and the revision after k )] TJ ET BT 33.000 718.794 Td /F1 9.0 Tf [(weeks. The difference from the weeks previous values after k-1 weeks is also shown.)] TJ ET 1.000 1.000 1.000 rg 26.250 409.165 555.000 300.297 re f 0.965 0.965 0.965 rg 27.000 696.433 30.500 12.280 re f 0.267 0.267 0.267 rg 26.625 708.337 30.875 0.750 re f 26.625 696.058 0.750 13.030 re f 0.965 0.965 0.965 rg 57.500 696.433 177.551 12.280 re f 0.267 0.267 0.267 rg 57.500 708.337 177.551 0.750 re f 0.271 0.267 0.267 rg BT 62.000 699.189 Td /F4 9.8 Tf [(Change From Final Report)] TJ ET 0.965 0.965 0.965 rg 235.050 696.433 71.925 12.280 re f 0.267 0.267 0.267 rg 235.050 708.337 71.925 0.750 re f 0.965 0.965 0.965 rg 306.975 696.433 201.601 12.280 re f 0.267 0.267 0.267 rg 306.975 708.337 201.601 0.750 re f 0.271 0.267 0.267 rg BT 311.475 699.189 Td /F4 9.8 Tf [(Change From Previous Report)] TJ ET 0.965 0.965 0.965 rg 508.575 696.433 71.925 12.280 re f 0.267 0.267 0.267 rg 508.575 708.337 72.300 0.750 re f 580.125 696.058 0.750 13.030 re f 26.625 696.058 31.250 0.750 re f 26.625 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 686.203 Td /F4 9.8 Tf [(k)] TJ ET 0.267 0.267 0.267 rg 57.125 696.058 178.301 0.750 re f 57.125 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 686.203 Td /F4 9.8 Tf [(MAD)] TJ ET 0.267 0.267 0.267 rg 234.675 696.058 72.675 0.750 re f 234.675 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 686.203 Td /F4 9.8 Tf [(MD)] TJ ET 0.267 0.267 0.267 rg 306.600 696.058 202.351 0.750 re f 306.600 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 686.203 Td /F4 9.8 Tf [(MAD)] TJ ET 0.267 0.267 0.267 rg 508.200 696.058 72.675 0.750 re f 508.200 680.176 0.750 16.631 re f 580.125 680.176 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 686.203 Td /F4 9.8 Tf [(MD)] TJ ET 0.267 0.267 0.267 rg 26.625 680.176 31.250 0.750 re f 26.625 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 670.321 Td /F1 9.8 Tf [(0)] TJ ET 0.267 0.267 0.267 rg 57.125 680.176 178.301 0.750 re f 57.125 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 670.321 Td /F1 9.8 Tf [(0.137)] TJ ET 0.267 0.267 0.267 rg 234.675 680.176 72.675 0.750 re f 234.675 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 670.321 Td /F1 9.8 Tf [(-0.030)] TJ ET 0.267 0.267 0.267 rg 306.600 680.176 202.351 0.750 re f 306.600 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 670.321 Td /F1 9.8 Tf [(n/a)] TJ ET 0.267 0.267 0.267 rg 508.200 680.176 72.675 0.750 re f 508.200 664.295 0.750 16.631 re f 580.125 664.295 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 670.321 Td /F1 9.8 Tf [(n/a)] TJ ET 0.267 0.267 0.267 rg 26.625 664.295 31.250 0.750 re f 26.625 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 654.440 Td /F1 9.8 Tf [(1)] TJ ET 0.267 0.267 0.267 rg 57.125 664.295 178.301 0.750 re f 57.125 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 654.440 Td /F1 9.8 Tf [(0.083)] TJ ET 0.267 0.267 0.267 rg 234.675 664.295 72.675 0.750 re f 234.675 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 654.440 Td /F1 9.8 Tf [(-0.002)] TJ ET 0.267 0.267 0.267 rg 306.600 664.295 202.351 0.750 re f 306.600 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 654.440 Td /F1 9.8 Tf [(0.101)] TJ ET 0.267 0.267 0.267 rg 508.200 664.295 72.675 0.750 re f 508.200 648.414 0.750 16.631 re f 580.125 648.414 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 654.440 Td /F1 9.8 Tf [(0.028)] TJ ET 0.267 0.267 0.267 rg 26.625 648.414 31.250 0.750 re f 26.625 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 638.559 Td /F1 9.8 Tf [(2)] TJ ET 0.267 0.267 0.267 rg 57.125 648.414 178.301 0.750 re f 57.125 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 638.559 Td /F1 9.8 Tf [(0.082)] TJ ET 0.267 0.267 0.267 rg 234.675 648.414 72.675 0.750 re f 234.675 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 638.559 Td /F1 9.8 Tf [(-0.008)] TJ ET 0.267 0.267 0.267 rg 306.600 648.414 202.351 0.750 re f 306.600 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 638.559 Td /F1 9.8 Tf [(0.073)] TJ ET 0.267 0.267 0.267 rg 508.200 648.414 72.675 0.750 re f 508.200 632.533 0.750 16.631 re f 580.125 632.533 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 638.559 Td /F1 9.8 Tf [(-0.007)] TJ ET 0.267 0.267 0.267 rg 26.625 632.533 31.250 0.750 re f 26.625 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 622.678 Td /F1 9.8 Tf [(3)] TJ ET 0.267 0.267 0.267 rg 57.125 632.533 178.301 0.750 re f 57.125 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 622.678 Td /F1 9.8 Tf [(0.060)] TJ ET 0.267 0.267 0.267 rg 234.675 632.533 72.675 0.750 re f 234.675 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 622.678 Td /F1 9.8 Tf [(0.003)] TJ ET 0.267 0.267 0.267 rg 306.600 632.533 202.351 0.750 re f 306.600 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 622.678 Td /F1 9.8 Tf [(0.050)] TJ ET 0.267 0.267 0.267 rg 508.200 632.533 72.675 0.750 re f 508.200 616.651 0.750 16.631 re f 580.125 616.651 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 622.678 Td /F1 9.8 Tf [(0.011)] TJ ET 0.267 0.267 0.267 rg 26.625 616.651 31.250 0.750 re f 26.625 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 31.875 606.796 Td /F1 9.8 Tf [(4)] TJ ET 0.267 0.267 0.267 rg 57.125 616.651 178.301 0.750 re f 57.125 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 62.375 606.796 Td /F1 9.8 Tf [(0.053)] TJ ET 0.267 0.267 0.267 rg 234.675 616.651 72.675 0.750 re f 234.675 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 239.925 606.796 Td /F1 9.8 Tf [(-0.001)] TJ ET 0.267 0.267 0.267 rg 306.600 616.651 202.351 0.750 re f 306.600 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 311.850 606.796 Td /F1 9.8 Tf [(0.027)] TJ ET 0.267 0.267 0.267 rg 508.200 616.651 72.675 0.750 re f 508.200 425.422 0.750 191.980 re f 580.125 425.422 0.750 191.980 re f 0.271 0.267 0.267 rg BT 513.450 606.796 Td /F1 9.8 Tf [(-0.004)] TJ ET 0.267 0.267 0.267 rg 26.625 425.422 31.250 0.750 re f 26.625 409.540 31.250 0.750 re f 26.625 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 31.875 415.567 Td /F1 9.8 Tf [(5)] TJ ET 0.267 0.267 0.267 rg 57.125 425.422 178.301 0.750 re f 57.125 409.540 178.301 0.750 re f 57.125 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 62.375 415.567 Td /F1 9.8 Tf [(0.048)] TJ ET 0.267 0.267 0.267 rg 234.675 425.422 72.675 0.750 re f 234.675 409.540 72.675 0.750 re f 234.675 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 239.925 415.567 Td /F1 9.8 Tf [(-0.004)] TJ ET 0.267 0.267 0.267 rg 306.600 425.422 202.351 0.750 re f 306.600 409.540 202.351 0.750 re f 306.600 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 311.850 415.567 Td /F1 9.8 Tf [(0.020)] TJ ET 0.267 0.267 0.267 rg 508.200 425.422 72.675 0.750 re f 508.200 409.540 72.675 0.750 re f 508.200 409.540 0.750 16.631 re f 580.125 409.540 0.750 16.631 re f 0.271 0.267 0.267 rg BT 513.450 415.567 Td /F1 9.8 Tf [(-0.003)] TJ ET BT 26.250 354.642 Td /F1 9.8 Tf [(Finally, we measured the extent to which ILINet revisions impact the efficacy of forecasts. In addition, we measured the )] TJ ET BT 26.250 342.737 Td /F1 9.8 Tf [(differences in a given weeks ILINet values for subsequent reports \(Table 3\). We found that the magnitude of these differences )] TJ ET BT 26.250 330.832 Td /F1 9.8 Tf [(is, on average, 12% of the standard deviation of the final values, with the average error gradually decreasing as more data )] TJ ET BT 26.250 318.927 Td /F1 9.8 Tf [(become available over time. Nevertheless, the most recently available ILINet data values are, on average, inaccurate.)] TJ ET 0.965 0.965 0.965 rg 26.250 82.642 555.000 226.405 re f 0.267 0.267 0.267 rg 26.250 309.046 m 581.250 309.046 l 581.250 308.296 l 26.250 308.296 l f q 450.000 0 0 192.750 35.250 106.546 cm /I50 Do Q q 35.250 82.642 537.000 17.905 re W n Q Q q 450.000 0 0 192.750 35.250 106.546 cm /I52 Do Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(5)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj 359 0 obj << /Type /Annot /Subtype /Link /A 360 0 R /Border [0 0 0] /H /I /Rect [ 35.2500 106.5465 485.2500 299.2965 ] >> endobj 360 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/files/2014/10/nowcast_errors.png) >> endobj 361 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 362 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 361 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 57370>> stream xw|TUMK'@Xa]AW,UUYVqU)* NM ғ7 äMIf&?grʹo޼=K!@  i@$MC !@ 4MÒd@ e(b Ж!!@ 4#$Bx@hӐQ@ ih(H"$H"ۦL@ *dh@ m6 0ӧOxOMMJĎ;B"B+ŃB,rɾ} DV0SO5jvZ4iV>AaB믁-{.,???%%%'''++Ch׮E]a%%%dgg'$$ݻcǎ]tsEQ=Z\\թS'Ç 222t颪jII$I{Kv{nn'233NGnl6i{q:YYY3\4܈v'$$3gl6[nni}u}.W^qqqx}3;޽{].,EQÇ^xطuرc:t 񎕕{ .oxq_e/pB!C27>?0%%%++{ڷo?iҤ̜9sȑv_~aÆ;55uʔ)oڵѣG-[@tݻw'&&nذG2eag޶m=o~홙?k6| 6̚5뮻믟x≑#GVTT 8:u(]t :}8mf͚k׮ݵk$I۶m۰aCJJ}t:zݽ{ovm޼y]t駟|/㙙˖- 'Nis\bE||Сݻc͇ '|g7o~3g\wu]v^w7n\^n^zftݻ/_ɓ_a(8n-wuc1ѣG9{'x4%K\q)2q[no͚5izΝ;YCYVVt:k;aNsϞ=={uݻ( 6o_^^nfiiiO8a۷o^{Ϳ}iiiFFt'N? ƍßo?4o'čK.>}:{͘14޽{oqРA뀯LKKK-֭6ln7or\YYYeee__g~ǿqSSS^+Cm@,!X`پ;V'v  U /oر~Q%⥗^o߾aÆ%''w !w뭷^z饃4iѣi/7o^\\˲eee,رcĈ<ӣGxӧv{NBvvԩS?QFM>=>>[nK 2iҤk6Sǟ322qٲeˮ]ƌ >xX2//߿?>>7xA:qQFmٲ%77n]O:(JNN?(.kn{РAW\qŌ3.ݻw?X+2''S{V1bĢE|UW͘1k׮$˪!DƢ6 )) ΂Yf[y^Qrl61n(jO>/2{lF?άY֮]]RRҧOf 4AdYƟUUej^tEؽ{?;>[pSO=/̚5kҥ^{m?{zjZM6sLE9w}7997b?eJKKׯ_?s{j9^_XA~s^y啦iv!~cǎ}7\sݻV_UՊ?$SSSlrO?tĈǎH"#+cx!BhǎVھ}ӫW޹sg#$a8ˑ#GxԨQoۡC.zLȑ#Ǐx,Yd̙(yÇg&++kŊO?tyyyJJ7`q>x⬬?3g0ڵkk?W_%g={ܹs?Ν;7++ u[u놝}&''Gss=np˗)]tђ%K7MsѢE'Odꫯ;v,0+W| EQ:u!C,XO̟?ܸq'O$ɓO>dee  ,СwQZZzW,O\y0x`-P5tPKJJ~+I9C$Xj>>/p6믟8qiJJJ.;wo>J!P@4"馛馛p&))iYYY0xo8C Bme˖=3o t]X,GIIIeee6m׿'G NUis=}" #VK<뭷N>8r{t]=eYѣG;6Z Y99yʔ)񊢬^zȐ!}n6m^}_0SPPsNRCkq&jJJ<i2dH׮]I*Y B 0Ç8zh<ZrmvmoībDQA ԁGZ$B iʕ+?sǓ=cƌ>}|6l禛nۗ.[lӦM4M8svj<_t:8n&Mׯڵk'M4iҤAաό3.2xWn݊⋠GHhM[lҥ 9sΝkZ 6lؐ9v澌%Vcǎ׺rׯ1bBox$<%..nĈ-5nLQTǎ;vX/G-ڵOܥSNMәշ!h2"4ò$I|#MOw8ݻw?y$'O}5'ZBnw3cX5rݦinݺ~߿?p w>q#&@ X0\rG}Eq޼yz~{۷/EQ˗/L"5#GtZ 4gΜx<6YRobXMiՇ5Mc٪i!k֬߿xb&3,\0ҊD#1ȍEQ5fQ@ m2pD 6 1V D!$=3ւNbxh@TU4_8: !@ 44-Z%&4!%Ν;MӌMleTMPbAI"];"lw': |Ѡy]lo>J 9ƥ_yPB .h[+m 5WrL7Y Hn:CbDN9k,<}5!2ʥ^$͜934BM$YfhЄȖx5YTuŧxBߞa]FQ(1ѠI$>A 4.{w Db 6TM 0a7>lr޼yaD&DFTuA1BY-Z GD& d!SPa&2M6L=ĽOh?d^24J .@ 9L'!lόfʔ)dyzB CĉU+B CÇѪUm<BBBS4MѡBجZZ S+DQ@" ,pEQ>2 C^[!J !TVV5 p6GhlV- 1U]6ny0"!QuY y*7o^jEhaOhVMZR* yM"(Ei> &sa({(mu}۶mv[xo,D*MPߏWhkk%7QjU/Lq}"(i ="Fy,D!gFqe7㖝j Vys,EG!ߙGbA%q j4MoٲE4m4E* ,"QO$^Qu( T " .Əd(Ҫ MbcxL(Q#k]Ru%JL4h IC׼(jwqǺunݺu۷.DG-&G(Fۦ?>0ag=ˆڤn'S<2 3M"'gP!- /^;hD& l5M9rpܹrbŊԶ4E* yJP4 yD^&~j ^t5o;R34-`38*Ϡtg aHL4hެ_zc㴭 IIQ,QUy8B}~];"9T]Fc;ֻfUG7T_yO(N9`TBMW$k4,D4W^}]wMKiY5h۶mÃM9n[GbΊkى Q֐5sE)6ꍺsΈZ 1°(-MQԄ &O_ׯ_~4x4S]420Fh't閝6!pz~Ohq<=+}2(0Pv~֨[5]эKk/Yz3~eY^bE{MtaXd@~~io㸷~[ YFiqUEo*=+80 Sx M!(`ҥ ,lIII Զѣ999 !k׮W_}!0L>}vޝxʍ t" #f:hC-ٹM!77SNM<p2|aqt[ɇf{\Ggj2C{v)TѺw4M3//c"ai(\)۵km۶^zu9///(q .<|!C$I(bU@t:l};:2 HLZGaG?hFp/nbp͇K.LKQ[G2iF`$ !FlP!%8}u(ҥK.\taiY-:}+/_ aUhyu[)&N24.pڴi4݀RWaQE~K!di 5S,rI[2P/YG_nTՔ V᝾D[Jm; 0yF+P|gۑCfhߨLxV uNz,wѢEGӴn޽]vopu麾o߾ ._ر#=޿޽WXQ[|*xh4#B!SF@$Ujj-4}|R5tСC}dY4 4 WIf0[bE3TЃBqeMGqɚD&+4CL PIxwIlFH1fh@xA?`Su0:x*oP# "'|Q&, -hX,@t#ڻw; #+ CR-/epQ``Y beMYAyy!o(.MW?nxV)+flhzȋ7n8, ,Xǎx;hYQr"O7{ɯ<BB(*>>}2a?4iK5/PTtOגj2n6U{0 @ywݏ6.ЂqÄB]WϹhή,Xǎ5&]'q͛76&`EbNTfBAjz@LTk*sD!~tCQe~`K@0<~6fn_e{fT5Df۽3Y-M= ` 1enIIkW8~GT #$CM-B>$ED}YRGl従[<ӊ !oA K.6ouG}LdaAS@N&^I%?Eb"5z<+24SwR2#lI2UR|P+6 ћ,QU_~|_~y||],(REIJx. k>DALN6qTtBMW3'g !C37}_FFeb"Yu ߷?L薝<+3OY9w˵&z$X--@h/IX'N V(:ë[e6?{:Q*ԼG_~DUGv򜈇z*Fԡ*'Ÿ:fPxdUKyP@,WGN$-##o _G1Bt6^)ȳ6GkU]*Ч1b5\lٲ_7|_|Eb>+RCxxu5DGZ {Dh E9gjU|yE[*BQѣjV:=B@R:L ?~K6NgBZb*;- :ˍ Ey:"34fs _[whq˻dQ{$Y%#GB& d5#leBǏ#09/=0D!%Ko˖-kJHQTV UgMR5%#c!E.Xj0zA#I?USQoGqYn}ůPл7Ur:lV1?7deG9pdG%lmoR .((*pt!{eWmn{Wàz;O[mH몠 c404Kn76aMӬ8:04ɆQs֨X?ɗU/HRX+ut;vH,?{p6#LqSowwE+J( 2&HO&TR n7X`8*6fjK7M3/// ΔEBr!ը*MQ).tڄ!5Oqĥ4 V=U/p, a\zySPI ae"J]"*0T=JN;]V*dT<'bg1B WShz)Ӟq,U= ԅy]6c$UyYYpX=1KS|A>u.5bTE@ uBNoA7pm!Թ$] l6l!%Dp*a":[HJ M }#lD3h<\iU`O.Pp !T4o⨬zl)ayV"LԴ@0xBWgi7h2DJ]IT,g/sL=8N r VBz3 çxq_Wel$j6nW5{Oendi (g:w>g) P, úZU=j@Z=³2n7T(-z,ᇍET[O:tDZ)S;${ 5 tS:*ѷ)ڄ!,sJ`-hhT$^=4MEl)ayV;_#TuEӕcuG74PuC-,hV${{pRU$8:upE \!T)&DzCcO5{e>X,Uu➢# U=xӄ !6ؤ `U]> h^E8q!Ye- Y_CQ/0vzwOH)#Oac/<Ы\ eiĪ!4M֭;tPZVV?գ֢"҅e4]Y޿lYCcG d$wyMk6%~t(((ݻ>6wܙ3g>Ž{2e DU!}2/R1B8QxtC,U7hSD0 <.J]*L~ >l 4FeXeb9Y ;ˡB^խJ5eM_|z\v)^dEcJJjH%VZTug޽2UdN%U$(**Y,(#ʿoQ3Jg,-ޚ,o'U\EAWzʒ*oBfj6ӧz=>dfBb#ߙ+Cs`w~!+#1gVho%;v?MXqbΕ;w̽5fؾ)9B RA,n$1{,zBCЗy)iǬ%]yY0&}J1Y.! CiOGC~4U6HdQN9W D5´(HNe77*2 ~ը2KqB89Q7zgWoVh{~wM*>OOVUJeYL!F)]^^nZԓOn=YPcF]>SD-+s?z}2s= /- -RBmZvWaĂ^֔^wAGyniU$#.MXV0Jm'K^%=mk<+H׷ܑ[+0>gJdy~MZ< 2D(I0{:9Ȓ˼vk/wT24 J?屍p-蕥*2ӝ/O<>뮃Z!"qnoocd1Q W7$($dr|F "걥_8 /_G}D@1d28:UNv.ϓ K՝#_lD#3!@b#t6_~ _u} 5ό"iK$E.EL$$$(  ((;whFg )5 mUI8Iݻyh{XkLѽYLSmYR,,Ðpir.BflVuIBID,ߧٖy- %Lyoܸ41!"2H:e'L,^?ڭ,=lo@+++oYWܻ&ʽ=>J)G}S눈'NdGϟ?9 _-<(ea6BYhz;cO^pJ)3I" nEKs#F'E6"i땓;举Vɽ#0Al!ns뭸 > "%, ◾0 }2ggٖA08q I:HQFQP JL쬘FEUgЉ:6 *kaDHD߶i%e^4KEҦWy4(&2M?-կ~ƍ a2S(,aLMgE$пMkw2|ap$=y^e?^w{1??'><#st) RJ'Y~#(Vs)8@U;,O5/am %o{ů!US3d cqN&0 (%aM%;6ʂ$͸mTj.qm@to9Uɯ0|S 8Z.2l U||YfVc֗+0ʯ,>2;Xr(ec ("QdRI{-/pN89~zG{%Zu}G?я~z~ oxOO_|衇$N*;k"L'c'% O ߈hEk zi>q *DdYБd,)Z])-Z>1tzw48*r܌67ߌ?OQ4:ݷ;*y{aDlIۇIDhUQt 0K>`YE$"\Dx`!ݺmWСۋhTԗRf!W+nf7> D<#`ʉcz<БVXov&& !K+,h]'qwG`J"ҮeXL8{.FIHT M%AZ^jC=|ѳg,Їӟ>syp%c% y?"<",v8}{μ\>ڻ#QuKc Gpܲ?Ã7Z9W\)mU2(e*pbk{$rN D8RETÈĵAQ_bGu%oGQ~?g(CF-MN-Bu߸q#orI1"e>EvA$Aa]Jn vWJR$("/_e~8}"ƒ)Ssd2˯6SZ^+n\ 曓F[Z@<Tť?M347A'w֤Z>Yc'l,<J)aeR#`^ 3pqD)Po";ǖJJ#G JӞ]48'"U#2 !"ql" ;; OLtzBnShv `۠i8e. ʂ&D>&˹0YzuZ=yGN BZ‚qlec?""dן BaٚAB!$AF_ oǍmQɯ "& "gUD58XXD8ݢV !T_b|X,pMBs3հ^6,QiNJ4boܸM|-)?'FFymL"B4N}7aL?Dxv>NBrF"BsjtzG7/pNgb4꺥y+Z v@R8/W*wsOV9k'qRDR:6t/:6="#,U޲(4ϋjjdkK)~򓟴<3gT qT>S.jajfFK$J|N2(PN6"TjvPJ)혍 T(ɵ BUUIM؝^ETlm0$) 7I>>\.7۩Y_ʯ,|/p8+EҀRC=tF Л#y'p`ú*障 (ciy>"L3l1J#ƘI!%Fe]o^ {ڟYl u9t|Tsqr"ء oo} jœhB@@^+Je~1wj{C'j`a.GYIk YZvTdI-FL[qv72FYTLL=SGCE908QbRw!I zǷѶ(2DH_%ۏcD^_0YT^̩Ůռpnm鲜ab#pmȖM>+Tܣ, fJ$Sۍ=7C_5 "xĉ0#IMe1ሰc6Z)կ0TG{g5>"DDn$tGo#7 ߒGhqE^BA "B7prx=;9؊8"/-Wwc[ͺ&i`QUA@HG\V& KlgE}ɘRվ`r|+cieeA"Ed!,K̶dL5 4矻3Weh? M6(e0fYH*s<Ї -f՛a#$p]/}/' GVXo=/RD82rR~W`En蹾"4y$i*BxdfEl"r(0(ogbQT˅U2dtJ[3H58 4߉: r+ZjqĠpw #mn6 K|SqQt㻎ij4 R \w$LYFeg9ZRB\Ҋ` 2 2Kq= h$ϲq8F w;qjV+|C-ؾ^xM-kf<CP59΋mY"X2H_#  oc_PQL1AnT|kD<'qM]?b=K6iD""K.rXZg`)첣On~?؞vUiYe GYCbiY(!0uu, eݰXK~0x~ 3v:nmQץC%čC;~aQN ͢ ,Ҵ^L77kTL"t,24ZrrB8o;/в}\n_OҒ "d߇%dh^؞95@@g~?_Q:ZGO?MS/L |QJiN-G78/ڦ>h8DjPCMo&ah "n8'4YnjoٸǯotD8 |"Cj#Ys6~DXO@QJ/\L)?wjRj(H eD|o5j9>yR@$IFl0(ahiqK v~Smbw&noSpmuQDyJ[om<'t]YZ-P(gPJzQ_9Rj"mbM ׷ K.ݸ6biFH#Dn~xs_ʯ}ckN6}/zzn`K8Bf󽱍v|-3QrZc6R z2ۃ >9z3闟?G,w}`ow&xcRnvv|lz=V ?#Z\gmYh#"Z$feF-y?7Ak$klB_EBEA]OчVbZM]1/,_dIP,opg7>7aAS a/Mˎl dkÏjy#5~2'S r!!IYDgB](d尌*'58[s9MQJLaƀ:=RJ2V,y Tr5xli#7/|fCd=j(ZQPX(wG>;t "LTJ,"7dVA?Ep}tww׭])׿݉67O!8Ƃ^\("=G@D`#" bY 8V63JM)5K9:7M2xJ) bQQ4'/p?9]I 4Uҭ"Wj|(E8GB:58( $q<^"4yQॡMms{bz] EAPչw.XU;fR bs¡vysȽn~J{W jѽ=u;PhGHf>e5NXmz j_P}eȬi04ɢ{8`O+n/77i>,o'5 uJ)} C9]Dr4>ЌeW٧ϴRJ^LҾZXk vxID˂K(eƏ?]-ς yYPD(IP+T9!L5z?od( Q6dЋ[ݹ_<094@1Cxl# BtL5~) BMۇp_EfME!K7yG~Hs]-llTNoO~&IP*h׊6p,>mFiQ,EQDUOq"V*m)z0iQFoٸOR4gھN "K<4u! -l)"9V7òhQ\:^={BKm0ki@ {!tWBj#DD]9('v&3"& YaKXBEF5g>6ܡu9Ǣ$=߉%7d_Go޸ry?t{o=׽ ؀b<{!Q w/QERIN|)2hGqw;bOK.T qXylB|W,%F3NyPsǗ>um( "6\|׃|; No"bO;Ε[OZ6܂!lo"%5[#b.n2rTIq]Gyg(GdΤA$y[ux~_p * T ugTn84 motvo?Fw.|{&CH#9"̮%9~/xZ!"iE[oN H8bu ")ZX)"@˂HPQDMôUe= M]"i! Lkv'ʛjt/&lI!J2;fl,s쯽CZee#/}#Z :$߿QC_61`T>TH5U0D;ʎFQ1uM6ʹvҼ |R <[ith,ۣ"xtqf#e0Ԃ(plNxK(b3H Ւ0@5FW?.C`.)q([ 8+g}8p,2_]R; h= r[ lQo4hg 3N_`rTIya9 Sf"2gR],",(~@ؽ*K\R쪲,2o' %+9jyTQEMo`(6r/~ؽ?pX,Bq(3COx~ziӸבkI9T$*lJ%X[ÝmwrjU_;tz{9;[m"PJK%LQ%pדnywr"bZoO! R'5b8P۽=HλJO<DgkEBxYxl{=Pn15⩍q;gQDZX;}Ǻ&3D*9/5ID(O f#vp~EA$Bx3!^3l0gty{a=BI5Լmcp8?Nx;ID;c){ր@1gp/]zsw܂S0n[ҍνCBe8AUܲ  `ҧ_b9eԪn}`0sv1(^:=@,'ɸ"a0p BD4Ԃ)aD扇 <)ǁ'G5^2G<4GP _>TK5R 1^9ݷ7W0""Xx}$Y(3{AA0i#S?,M7JVVq=̖ W,?rbCY*;{u03 @=WJf:e-׷NdY՚-HN)t\Μ/~q,bg+!@R]y^3jؾ[DZN-{P.L jqj"/6z䰅0PJK%pEi O>(eB&EN4rEjwH0*ZtuW' g*ZXN9M`cN ",/'PC{:[YBe (UY}Cp+NOTozmL BB0Cm0 %โF3eJjRʊ5Vr+$ NQj\XS%0|^D-o $Ȳ1A7/|{w#밲t̙6B8t[ޓloFT5jkNZ ; IL0 a12҃# P@z#VA}Ib-mcz TZyVRU]Q ,@,sBAp]P:o>W_su&vAg(KJ=~ ήW} "8 !ٽ sEDߧ&0AP,"?͡Tv,n GfyrB9!в!B/pOdP ""&TʂCl!<j^osgq~3+eq"6B0 4^׃R e Lkﭷ߇bqqs؅oYКD% x^D86œZ܏G8LwbrF !&;h55( %d[ª %c5;tTzW˹eYPzow\ۣ7`~"/Kl{s+ufzbuaϞf+tΰΤcDX,"qsJz!yNPD<{^Q˜QHB 1|>qw-;1ͯV̳Tpw~ZV S%5̶l6)a3vikr"Ta t0)4@}Ge$Lg&sal*+_ILw}~ \3nӥ%؀xz,8C^9UW~Mozӽ;}>vmwyLG gmq JonN-C\rd߲~m`ceHD<P`;_V$b{V|uiZ QN i\&7K4zuW'꽭88W[oZ33u"ht:W~hYP*vdU%ԊĚr ^4;i#A@i$00-¡p@UE1E g!\9QvF˅pЧ\-@Ģ^@g0dxIDA#8 Jcil]Y58i CZFb+!ps"B "CpMuE0HX"OX s尴2Zmn)hne,kyMo>}/(ٖ&u`%\<ÀZUnCTU1aDl/;j=G6 `4 $ gܤz?6?x?#`%'߽ @KKq_u}-oʡbD78S?G?wFLZ@ 'k#D'pw\˅jaلr9;:[[4&X.U9#mۨ54`?"m焮dGE)`,r~}"Fkt zowmCjPo%n,kn9R]H;fcZ 8O;<e鶗N3E!%~9hÈDqH" ,DKYFIRAB&ҤT⋶- h YmI{o7[/d[6)\ZWpJVɧa-mł0oFEDET@DUE^ fv̺[[ԶJk5n;~Dj;QEgHt.CV"N"C_ňg~gN6=j__4ǔeE%[9QC.u^$Fyj5dGJjō( f}/pcdRA\Y}o)-pXL= BscNE v|K{"k=|a+ܸ! 0_)!BXZv{44"^/q9ތ7V4AAׁێͩppTHdF6,b~8e.% ˅a;Y-^z2?͝q>*_{"\ QvA-ϢdOuYT۴PzW,2$(!7BɓҠ2D x^,~Vzbo\TTe7 Zn='AF|O+W>0NW`yIDeqlCzefʾ.|JPDC"_!"t)6;6nw/4nay)<_Yl[dvrN>ON8qUsy00"P"(Ձr dI^ q}=ANmi(^iώ&wtEE~|GcE^++4WշͿI&U !AXR,bj C!q^o}ڝY!sYm?4qN7Mc`eۣ?<7+4ڷ{ CIљ0 $7=3~!lTrY.!<ZJq,W{f+dvV ze`w X[jni+_=woZ&x;DV֏;fg}qgj5 '#輳8rN}9=DҍΓct߯a$ g|mr@pcC~?B'W,5Z'ʕ\6x 2Bo&s0pla"4B|Gݞ>oKze?mcET'}NK<~zNwLoO6T;ށ<$5Mur J_ňpzEX"b x@@$ D K)elԊٖV9xC\PK88"d|4( ]P Zs*B'e#ҵ 8z@[ӣ2F !Ge=qre"BD8}/^ܿ'@H=W+8ب*imzltd}wZIz{jax(F++FJn(&X0 ޹;ۋYd^waQDYΦsxxqh47-Ù3\qxsnL5:Ge] k ,/c=]2$AoK(ju?!dhv}2n UEn5 8#"9TٙJ{O<Ƿ ze*-`c#Ar<~+ +r}7@snr/an4B;;;?~{o-4 P"B7]` ~h#:Z@mvɝ3[ 3<nXs~yqaĕE-{4Ev" 9"F^?2Z:1i˅(KO^kaVE>J3=y776@?Fp4\a "Dߓ,o{R9nR~ BLM3YYU6 %p~rЕ<+\]~GOw,hm8ۣ]SšO<Ƿim ,qע~u|xN '= 4c "xPΝ~s'csaC,GԛyV'+FrvKw~۟_Gֳ;ckAlPRb˶B!'MD,].)KrjNշ۲1ڂ%8v i%W :JisR:V>q~i8x6|InM4*Fm[E),qx׻z zx=e- U۝LiF7tm Ck ]ZίNM_}E/n yr糆WZWWF!/כ^}_}=kRIIx_Fk&WvG5E&[[qϾV% N\mcS/\}ߎ n/+r/< OO|{~ԩSƵ0ը<_sqrg[&q6JVHE $dd@#;sc֙ g_F,"ygvvrf{G3#$=g~}5izpG>JuZ[,LBckBdE!Qv d7Q% t: s >#M)I("q<Aa<)8@ ID1RJux]+Pq{Dp ǃUi0pє\٠0>b`2 ZY!$C a~(+±@r0"sīODO镩,"*b0'x>)ҥ%TBi}>bA+yTdk8*GvC2 ǣ>p J+B&*@ssji_+@7L?4}Uo w?p(&tD#$@D"d!̃snu˭a`D#N' 2QqJ Ev].< (U@diɢnX> n4y9HL[Oq`1I]nL6D#1ȶg6H090op@%04!ꂲ2d2!b hu25'Sں< qSЅ/w7;_|Q!ۿ#XБ#t!eі#nP$au+FQ!Tiou[~^Odk4=B2I& >!adׇʡڱDEhz_iQ #D~@ɱ? 8a̫ȉb}ŦGwNٌ3b4JTv^sdrAi)x< [1T(:ЯF#Tƥ$(` ٟ8HDcQ#~0 `CKq #jѱ^IA8LV`24W-La $D,"AB(S'x2&0ۋf(+C{AGMQ=%$غȿLVL^$Z !P(`BOޒdS@*%GlRH{ExhF *+ezdx1@@l9  Od6#HUNԄ P _Ēf3jlSwR;ۋy!"2 Dʴo`(y]"nAbq`^e0`ti >\V0&Iz7VWz%:<"]y'2q̘A= Owa!˲A0L !$˕Je*X$J3Fnh#z=rqͮ_iksɖGbDTΊӁ}(6.bRF^Ǭ)I I!`Y‰Le2PȩdTt1eBN6wX,DZEc)*3.S;bG1HSVHVBR::QN0U."H A*WJ%Ǥ Rq(ˡ6UHZhvZdȾk"X">OFW{DL8,rJXbЇ"doPBe`xP@r@00VrMxF|4ɧ1S>H? }W k*1^BVU%[Zhc0ZѮ.P*sۘBfs)4'RiBd!q*ee5(Bh+ ߾.J%b%Zw8^0ǂ,Z,I$B,Q(U I?pv~R!1zo5Ǣ)\G'8u8W] &l:QL?sﰍcl 'X|\Ƙ~F8BkȢDDb{zlR6(_$NऌlVaIDZDt0aC1 ͦQ5qXP2&˙Q@dLd CKF< CâXbz*}l8 /7Xx'ݘ0D*2 v|w;r>R"D\^X8Hd /BPXw@H玿rˑ}aNvgxʈG4őxhlp}"]9JhBbNNK9 sP\ &B@L`* l <Ė1E~ZXqhD_nՠ6qU,&"r8 YzYZn>~dѠlhḁ 8@:CSocԄ VP"Ak{#G:DyףdT 8ҏ bH wB.N$pz b#TH8sGyC:ΘZ,`2A$osz*R˟"{-R)ZE)T \B߻̩3Od uH'eo)FD< l&@K$A8 Id4A4;3PBHʴC[bTP`{:G1Df= *7v85˜\Z$4vO#AfYQ 8±@yT 2=HC6J(dfK'LYoԿ"̈P*! %af!S6ɨ.AY~\qRydi%!U8p(U„rJJRy<4s25Ԗ`GX[HD!A9ͰlYjpt4$sp$.$d#n@qYupujp<ͽOhl 2p,eCR{$W-<}ru4${mGu\F'D25Bv 9"eL :EyRH#~ҘHW.-@DRij% -Q+tĤY8f3gR)~( ^,R !+K IrVɵ x$iBCѪS0nΐ KeF͎j;8(5b0C $HiqJ5Bɘp% ̛ wLVR r]HG]Ч۽Dn#AQH9 >FH8 ` f~DR= 2BYF%`YɆu.x͆p}n>_֦aeZޮ q_nVch0Ä! 냢"AʠF!A1vlIsAz+󼬇V!ɉP8?d~ ^5 ۚ\,_-h:1ǁQS7-NExYC 0dXjYi ^8C )`[$Z*S#\w &Sq4X}!w<%+B)9l`ͅL؟JE!)KloYSLQ бþ}$(?% c(h'߮6Q3 d ׇe2&=Lr9c[p+F1v*0P=xC, P*mP65H41Fٳ$&u1jiɬctZ%TZ ,FR%('+^/H$Ĺ4DZ6rav>FH~ II"P(P!8>GDfvI?bm""jGL8]s鍯\:pMMDmT5P62 {,F|& OFSBB(Ƅ5!d0ƙD*%t ch#~ +0Z$S'Tj@F '^J!R"M31T 2|u:@9 h5Lk'~19E"XNi"!cd2N0 L̝m@[4 E qxB#!4cu@j!#[,(6BZcLHa@At֌xС!ۍZPUp׮c~Jb8;X WZ|!;/|"g z,xR *:y\ID>-1Ʊxe$idjlfYR'r"@.Q濟caJ mgswJhq({x beR (7˅=Z-+H6ncˆRZJ<>!`&aAZpP#ġd@QHMݎ zꩧfΗ W# 0 '>[;z+S~q_p|*_B5?c-<\B"ƅ)˥ [?LgS 0,IZs#*($cߟ)MiK4 Áf*acڅ?4 Q>G#(y|tUi3rzP@43XR @aW~Sdjb3eeamVw_ۢ-/V j#$|{ 'OL+ss|J !22O&Y[1A.tCfy)KX} :昍WRQSDuDca^5բDTmՕsEn7` `kDq6ӨgĀt4bVF c|}ۑ]WNe !B`_S&WnթMu V9HbT$s2UH+jNBؽy] X!GKYmKþX1N'X7S+r,pX"Pt:&f?&a8~A!hjg q!D:=Nd# RVKFnxPBRc8v /\HE  ٫}qݍ#~^N(/GvCeO=]d-ؽ0Sϼ:zqBp#yk841Dv}-gpƁ@G"87nV +$e Hl#HՅɖ()7ʵm"$0mVoތˁ$(pԃcƜ$^ϰ*egd"(+*cfC==H"P( ˂LJ}c aLDqB -JKY_L`єm`CQpF D!UԯLEF{{qq1`*\4% *7w2JD}`R)R*-~ …czEML!IYH,fIb3e4Y C!lFr2'8ѷml$0]R)z1cS~/ţ &;YZ}\[M@̄@:;qi)Bѫz=$υ3TC; " шJ\׶e-h9ֻgu+F^$8&D Bm-ÇSG2V!FVW#/nD/ŜLmڗؽ+hzRjTeb?@Ujccxm]fDldTctu%`I? 2[Zk,7Hm8MIzm먒 `HkGb&%˰|F,㮮h 8GDlKd2 5cF.A;7sP.`VbH<*|6BR# cp@(-gh3 jK3G6AȦ?+3yMQñ0'J7ǎaz7`{ޚR,XB--f v;TZ??>O'[[ .OMB9xԌF$cRpKnpt: ٚĀBzTOM܎ ?h#laFIQۇ l!DrRj9c"02 #-.F$!TdKP\Bà!wK9T9le2&B2)'e#D XXjz)+2O|@FlRmj500E7lIQ`2Ar9%h,w<-g"x,P(c?!d0hHu{j#ap6لG_p$ϯ&Ӄˇ$d.¨&E[&JF]n3Uz_d>qrFKD}CP ebjna`X ` evk55V\UL&(-{s'V^pdjNN$k0\ЀkOٌ@Uiq5 (] G+.DǙz)H\? #^a0v,2=㈥Jx$;Rӆa((+C1D@ ]VH`} eGŌ%[VH`CQEc,e x׫Z#hҘE>w[7 lēx2iNr>u M={eErb&0b&R4CS,0̉i"ϒc| TT I|=6B܃z1Ơ4BȗUh,ҏwKIX12I,jk_yŚ`0 UhKD kj8ISܛr;qrFyޠr|> PSlm% goM jn>N҂kjEĉh.\]q- v8FJBF?1bcQw7xCmv} j 2LZуRkL0,ahvZ<QP<'Sm2r*ҕz: EBh<5jD z^`( p . MRVWluRdXFr?_:_&]Me4^e޿W"Bk Rj8O: b1()9؜2#Dei9z54(tvQW#hdjnCTX aj(+C۝nZ[1)I2;;S4'Gq8OAeai8S Y2 dt s2nŲuwc'!*WD+$+XX V0,JzK8x^^~@)+ i4P@_`YF*a~{+Z# &"hO6*mޝ͟a#  $^O=qShP#%"B-D CE$'TQ,J`I$]])vϒ~aʔԝKa$L%Í7R BJe{ :ƘB+m<;7O3lcuJSgKь qh w,{ܱe-5arbHQ ~MqC5 rgZ^JDקDZDdU 9cJK-~u˦xǍCqĝ;V㛑O↮f4qZatӁ74K U AnO$tx!BȢDCryaJJa2 --9HYnU%Y!>_p"lv'T*8-hD?ů@n7\LŲ"hBwƍ(NorRkr5BH0BhD# >x*Պ| D|- fooKD\N:JgBIҤ)JO]/&D$Uh B$F4aFbCw[%azK)BǁҨ4W9 騦3+ Ɇ~$!9m {JL1c[Z0 jL8KK.B¤L Lej r2#*9mt:H&!VL2رq)& Ғ~lm&SjeawofOpTEv8xLD{E;tJLUjoJ$ȣb*5PT3_;I9Ch m*ҕxv~+6t}fmNҠakcQR &hLRVNQ4W#h@.;S!vA \24Rۥk,r c޿?2ZdJ7Lhƽ5NduNSioo)m2G^o*(Y>W QՖ1-}O~k70 RMMvH>f zܙ`BرԷ|K *<ލ@駅ѬY1aMIjo}}l6@uHMyt))$BR:!˪cUyPF xT !Dr<BC[qwf w ?{ Z0>S^\ Ɏ[E%%pfTBR*b:#8|o+VP 2j:!`NJ/+=m֭djW\ ~rRL)RAooj_}[*/GC3t=_ȭ:hzR #XǕO%FW{=>?H$H0\)c?g<9X9E0YNPSir֩-8F򇒆)$,Ƭ-͐@s3 Qqf;pz TWe5e $?A a\_uu0469)zW#%LS4c٦:]w9xBls|4]-,' .|Qyo&DL\Ne(b˃ٳQe%zcp:!oA`}޽ˎ땔Qt5g7Z 3^-„ .]h[ټY3eg4 |\^/|oT`rBr.RVFS41TCG҂G:iB!LsPG"ۋR0W^n"`YoަI7iD2v.m)EMp onm>\DXm^)lLr]__{5WwIZ  P_wkZfCMM B@BBƲ}&W^RK ph:A\ᜟ&j5zZ"/g_[.s 2,.̅ a現?^n%\ʧrѓN\@} }7" 9M F4 Fm݊w:;qI #,C07cS"$W:)]([n ;f vp=|$-}_ssՅ6En'RS m91))ZAg{zt([~ B4i-~:ߏs {}åD*鄁r'qЛo SSG$KPf@ˌd]tM$RG.QM%fPBHRlIƍC}}rAk+TT Yub W6O T*G?ylFIJ0 N 144 /YBeE;1Q{LXn,qB־< ͻ81Vdg[KSZ]pn,#[' ~szjg瞣ozec!W)!R/RVN!I#2t焖!B`6`/\2V섷Ӧ{索 ?cQE1_qغB]NiȦ@5ٜ޸:[o rH4¼A _*EK( 2cέ--OqIq0ƵH*Eyr(.L`dRj>jjB?IA0mZO xj5KԴG}մieFCCV+h4؈3CKB;flLf\zj]EBÞ˲2DP=^-,[F]ye;#/I=1Kg9I8Nw5f1N|J)eaBE uǩєACp]5P[ᅬ-;etydexR̜yb#D#ܲE8\dP|BMcx]A"7 &I(aw7߅?PSӦQ8I/K ܜJkI}!)cFGM衇JryAGOF,h{sh9Q~=%Kۓx<ǩ UN(]]xF.hno`en{%#B.+؉B`̘AgorU 'g lꛆR ho[oEz !rrPѣmތag<?q"=[t9sЌv_|!ʏ3p V !PVZtY:WA(/GwA]}55zMI>n :hY^!5*"N|64ݻqg簞2k\k=8Zg\f3('\$!*7R2ʕԒ%d$sWG>JكyO$ ^NEƝ|7O&G{?}T*˘G_TxsAOMpBA$mӮ&4grDQqPWr KKJBp>zB! 9a6M ޚX2 F^K=t}71 a|ce᪫N(T?Ɓ9ImTef*?Q?(#8. *O4 Q8|GP};O<~G3̩жЯ~E?? ӧS'̧=s@XA}bTQMܔ(-HؓZhէS@ !:IBơݻq$"qNL pT:ľs^ihw݂ ~9a6C}qQbP^~^( jtčߒ2]UeFZ~5TV:V+`t9gfHcc9jX\vulEɓǩZ ѣ.D$[Ʉ_Gj MMMv/2׃?5eԼ@1q"zaj3k N|UVm޼y&iŊ<ϟh :%`6^r9%@M0rOf0)E*g~hov LH:3GCW^-/A4oh/˷݋L<(=}zjW;X 551D"EEp\y%N$ ?œdw*ʯ{XxzG`uN%xO$cfҒ6UOxDW1˞pH&Lg ԙQQF555z3gwmX^|'475A['Nmw[ZN&Hr ôiӦMv,,,d?݇|H&v]Ӊ`۶m6 At֮][ZZZ__ܼjժ{W7H%wǍFhv\PHPh4t6qgqjF`H~mmmVU܁=z`08m۶;vL(cXwwwoooz;)Hhv\,.].׶mU*w>r4>6mڄҧ {<'Hofs ?zEN=cÇ+a4S{D6mnX ۵k7{8Hk4UՊ3m޼l6_fssFpꌭTh O9%KTWW,K<_0ޘ1cnE3G*//Z+Wӟdt:O?}Z,+***+++**9s\{˖-;wn0$򎎎iӦ,_?f۲e XO0eʔ{BѼD[WWlٲolƍ4~gD;3z{g̘1({_y .['O|x<1nmm5jԭ:s{G<wF+++l61ްa%\rm]z饓'Ov: B49sҥK.]:{h41~',YrwyK.bG}bŊΙ3g?l6ێ;0ƉDK/]d?|q| &b˗WWWK$gtࢋ.2 ntL_nw0,))BC AX|c=F䍍s|7۷dq!x<&,GD ƍD`$YjUB 7pv!*+++;z(in>tPBHc֬Y+~<<`^x9F1;~+z.o9sL^lٓO>IyWcxdEEEBO|`0fqIT*–-[͛R{͛  ڵk̙9ǰ8>;;;G}:/|DlڴiܹEQ5w/w^Zm27o<}tq?3{lr͛Gu]~j,w}o?LhFW}ݻwtk׮%ه0 ?ꫯ>&wygɒ%98 ```0F8 644ѣ>!e͚5}GGի \.BP(ȿz`^qb5A|s@ `4rPL&2 w H]w^lc1`͚5Lovhoo+} i?3Zwtҳeeeh---moogpt??|7c90d@ X__1O̙\!af͚wr}"cՑ)bY~ԨQGyw]xx 0jzѢE{:#G;S5nW_} 7nF ]f6~i4ݲe֭[\s͊+Nƍ= ׯ_p!9{ƍNß|I4---EYp EV"7ndΝm۶m޼9}g 6dC.\`Z?۷K7,)H6˲W^y?_/6"}'x;ܹsƍB+V+x MZ[[W_}5yqu׽۶m۶m[o|rxWw/߿kiiQ*ݻaÆ۷袋`ҥ|gϞ^zoڴi`Qr# ۶mۺuG}h>Cb<=8m6mڶm… ?#z{ 455yǓkijjZfMMMn'Lׯ_jժf ì]]f L5kyfm͚5~͚5ƍ;Mx:oz߾}wjA@dYYYJh4&%%l+W{* @iiiNN,Y^^766pj5ӧO|>_G<uJ0gǏK,Lp\6h,X]yJUUUpÇ=HxҺuܹ%I]]<m;BVrrr]]]DDtMMMw0  X,Rn2 hܸqcCCÙSSSW^-Hlɇ'͛7J%r\& fgpjuBB|*+ Z-㽽yyyA {zv'IRѨT2 zzz^|aÆ"l"ȴ /^,(JNNvlE1}y"ƍ 4/,,lG~:ɜ"˭Vkvv661mxxX*I<<K /Sb v;̴SRRRrss1 KJJMEr6 3}|Џ8 ‹/|8͛/]{??aNh"@PZZ8F!jd2Y|||FFInwccH$hL&jRvGGG].Jw~;;;VdUWWL&QYY ˻nxQommAAAjjz{{K$Bt:t`0?(#ȯCih]v PVZcx֬_$L$Y,T*rѣGp_|yyyyjaaa``3g> endobj 364 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/files/2014/10/nowcast_errors.png) >> endobj 365 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 366 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 365 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 57370>> stream xw|TUMK'@Xa]AW,UUYVqU)* NM ғ7 äMIf&?grʹo޼=K!@  i@$MC !@ 4MÒd@ e(b Ж!!@ 4#$Bx@hӐQ@ ih(H"$H"ۦL@ *dh@ m6 0ӧOxOMMJĎ;B"B+ŃB,rɾ} DV0SO5jvZ4iV>AaB믁-{.,???%%%'''++Ch׮E]a%%%dgg'$$ݻcǎ]tsEQ=Z\\թS'Ç 222t颪jII$I{Kv{nn'233NGnl6i{q:YYY3\4܈v'$$3gl6[nni}u}.W^qqqx}3;޽{].,EQÇ^xطuرc:t 񎕕{ .oxq_e/pB!C27>?0%%%++{ڷo?iҤ̜9sȑv_~aÆ;55uʔ)oڵѣG-[@tݻw'&&nذG2eag޶m=o~홙?k6| 6̚5뮻믟x≑#GVTT 8:u(]t :}8mf͚k׮ݵk$I۶m۰aCJJ}t:zݽ{ovm޼y]t駟|/㙙˖- 'Nis\bE||Сݻc͇ '|g7o~3g\wu]v^w7n\^n^zftݻ/_ɓ_a(8n-wuc1ѣG9{'x4%K\q)2q[no͚5izΝ;YCYVVt:k;aNsϞ=={uݻ( 6o_^^nfiiiO8a۷o^{Ϳ}iiiFFt'N? ƍßo?4o'čK.>}:{͘14޽{oqРA뀯LKKK-֭6ln7or\YYYeee__g~ǿqSSS^+Cm@,!X`پ;V'v  U /oر~Q%⥗^o߾aÆ%''w !w뭷^z饃4iѣi/7o^\\˲eee,رcĈ<ӣGxӧv{NBvvԩS?QFM>=>>[nK 2iҤk6Sǟ322qٲeˮ]ƌ >xX2//߿?>>7xA:qQFmٲ%77n]O:(JNN?(.kn{РAW\qŌ3.ݻw?X+2''S{V1bĢE|UW͘1k׮$˪!DƢ6 )) ΂Yf[y^Qrl61n(jO>/2{lF?άY֮]]RRҧOf 4AdYƟUUej^tEؽ{?;>[pSO=/̚5kҥ^{m?{zjZM6sLE9w}7997b?eJKKׯ_?s{j9^_XA~s^y啦iv!~cǎ}7\sݻV_UՊ?$SSSlrO?tĈǎH"#+cx!BhǎVھ}ӫW޹sg#$a8ˑ#GxԨQoۡC.zLȑ#Ǐx,Yd̙(yÇg&++kŊO?tyyyJJ7`q>x⬬?3g0ڵkk?W_%g={ܹs?Ν;7++ u[u놝}&''Gss=np˗)]tђ%K7MsѢE'Odꫯ;v,0+W| EQ:u!C,XO̟?ܸq'O$ɓO>dee  ,СwQZZzW,O\y0x`-P5tPKJJ~+I9C$Xj>>/p6믟8qiJJJ.;wo>J!P@4"馛馛p&))iYYY0xo8C Bme˖=3o t]X,GIIIeee6m׿'G NUis=}" #VK<뭷N>8r{t]=eYѣG;6Z Y99yʔ)񊢬^zȐ!}n6m^}_0SPPsNRCkq&jJJ<i2dH׮]I*Y B 0Ç8zh<ZrmvmoībDQA ԁGZ$B iʕ+?sǓ=cƌ>}|6l禛nۗ.[lӦM4M8svj<_t:8n&Mׯڵk'M4iҤAաό3.2xWn݊⋠GHhM[lҥ 9sΝkZ 6lؐ9v澌%Vcǎ׺rׯ1bBox$<%..nĈ-5nLQTǎ;vX/G-ڵOܥSNMәշ!h2"4ò$I|#MOw8ݻw?y$'O}5'ZBnw3cX5rݦinݺ~߿?p w>q#&@ X0\rG}Eq޼yz~{۷/EQ˗/L"5#GtZ 4gΜx<6YRobXMiՇ5Mc٪i!k֬߿xb&3,\0ҊD#1ȍEQ5fQ@ m2pD 6 1V D!$=3ւNbxh@TU4_8: !@ 44-Z%&4!%Ν;MӌMleTMPbAI"];"lw': |Ѡy]lo>J 9ƥ_yPB .h[+m 5WrL7Y Hn:CbDN9k,<}5!2ʥ^$͜934BM$YfhЄȖx5YTuŧxBߞa]FQ(1ѠI$>A 4.{w Db 6TM 0a7>lr޼yaD&DFTuA1BY-Z GD& d!SPa&2M6L=ĽOh?d^24J .@ 9L'!lόfʔ)dyzB CĉU+B CÇѪUm<BBBS4MѡBجZZ S+DQ@" ,pEQ>2 C^[!J !TVV5 p6GhlV- 1U]6ny0"!QuY y*7o^jEhaOhVMZR* yM"(Ei> &sa({(mu}۶mv[xo,D*MPߏWhkk%7QjU/Lq}"(i ="Fy,D!gFqe7㖝j Vys,EG!ߙGbA%q j4MoٲE4m4E* ,"QO$^Qu( T " .Əd(Ҫ MbcxL(Q#k]Ru%JL4h IC׼(jwqǺunݺu۷.DG-&G(Fۦ?>0ag=ˆڤn'S<2 3M"'gP!- /^;hD& l5M9rpܹrbŊԶ4E* yJP4 yD^&~j ^t5o;R34-`38*Ϡtg aHL4hެ_zc㴭 IIQ,QUy8B}~];"9T]Fc;ֻfUG7T_yO(N9`TBMW$k4,D4W^}]wMKiY5h۶mÃM9n[GbΊkى Q֐5sE)6ꍺsΈZ 1°(-MQԄ &O_ׯ_~4x4S]420Fh't閝6!pz~Ohq<=+}2(0Pv~֨[5]эKk/Yz3~eY^bE{MtaXd@~~io㸷~[ YFiqUEo*=+80 Sx M!(`ҥ ,lIII Զѣ999 !k׮W_}!0L>}vޝxʍ t" #f:hC-ٹM!77SNM<p2|aqt[ɇf{\Ggj2C{v)TѺw4M3//c"ai(\)۵km۶^zu9///(q .<|!C$I(bU@t:l};:2 HLZGaG?hFp/nbp͇K.LKQ[G2iF`$ !FlP!%8}u(ҥK.\taiY-:}+/_ aUhyu[)&N24.pڴi4݀RWaQE~K!di 5S,rI[2P/YG_nTՔ V᝾D[Jm; 0yF+P|gۑCfhߨLxV uNz,wѢEGӴn޽]vopu麾o߾ ._ر#=޿޽WXQ[|*xh4#B!SF@$Ujj-4}|R5tСC}dY4 4 WIf0[bE3TЃBqeMGqɚD&+4CL PIxwIlFH1fh@xA?`Su0:x*oP# "'|Q&, -hX,@t#ڻw; #+ CR-/epQ``Y beMYAyy!o(.MW?nxV)+flhzȋ7n8, ,Xǎx;hYQr"O7{ɯ<BB(*>>}2a?4iK5/PTtOגj2n6U{0 @ywݏ6.ЂqÄB]WϹhή,Xǎ5&]'q͛76&`EbNTfBAjz@LTk*sD!~tCQe~`K@0<~6fn_e{fT5Df۽3Y-M= ` 1enIIkW8~GT #$CM-B>$ED}YRGl従[<ӊ !oA K.6ouG}LdaAS@N&^I%?Eb"5z<+24SwR2#lI2UR|P+6 ћ,QU_~|_~y||],(REIJx. k>DALN6qTtBMW3'g !C37}_FFeb"Yu ߷?L薝<+3OY9w˵&z$X--@h/IX'N V(:ë[e6?{:Q*ԼG_~DUGv򜈇z*Fԡ*'Ÿ:fPxdUKyP@,WGN$-##o _G1Bt6^)ȳ6GkU]*Ч1b5\lٲ_7|_|Eb>+RCxxu5DGZ {Dh E9gjU|yE[*BQѣjV:=B@R:L ?~K6NgBZb*;- :ˍ Ey:"34fs _[whq˻dQ{$Y%#GB& d5#leBǏ#09/=0D!%Ko˖-kJHQTV UgMR5%#c!E.Xj0zA#I?USQoGqYn}ůPл7Ur:lV1?7deG9pdG%lmoR .((*pt!{eWmn{Wàz;O[mH몠 c404Kn76aMӬ8:04ɆQs֨X?ɗU/HRX+ut;vH,?{p6#LqSowwE+J( 2&HO&TR n7X`8*6fjK7M3/// ΔEBr!ը*MQ).tڄ!5Oqĥ4 V=U/p, a\zySPI ae"J]"*0T=JN;]V*dT<'bg1B WShz)Ӟq,U= ԅy]6c$UyYYpX=1KS|A>u.5bTE@ uBNoA7pm!Թ$] l6l!%Dp*a":[HJ M }#lD3h<\iU`O.Pp !T4o⨬zl)ayV"LԴ@0xBWgi7h2DJ]IT,g/sL=8N r VBz3 çxq_Wel$j6nW5{Oendi (g:w>g) P, úZU=j@Z=³2n7T(-z,ᇍET[O:tDZ)S;${ 5 tS:*ѷ)ڄ!,sJ`-hhT$^=4MEl)ayV;_#TuEӕcuG74PuC-,hV${{pRU$8:upE \!T)&DzCcO5{e>X,Uu➢# U=xӄ !6ؤ `U]> h^E8q!Ye- Y_CQ/0vzwOH)#Oac/<Ы\ eiĪ!4M֭;tPZVV?գ֢"҅e4]Y޿lYCcG d$wyMk6%~t(((ݻ>6wܙ3g>Ž{2e DU!}2/R1B8QxtC,U7hSD0 <.J]*L~ >l 4FeXeb9Y ;ˡB^խJ5eM_|z\v)^dEcJJjH%VZTug޽2UdN%U$(**Y,(#ʿoQ3Jg,-ޚ,o'U\EAWzʒ*oBfj6ӧz=>dfBb#ߙ+Cs`w~!+#1gVho%;v?MXqbΕ;w̽5fؾ)9B RA,n$1{,zBCЗy)iǬ%]yY0&}J1Y.! CiOGC~4U6HdQN9W D5´(HNe77*2 ~ը2KqB89Q7zgWoVh{~wM*>OOVUJeYL!F)]^^nZԓOn=YPcF]>SD-+s?z}2s= /- -RBmZvWaĂ^֔^wAGyniU$#.MXV0Jm'K^%=mk<+H׷ܑ[+0>gJdy~MZ< 2D(I0{:9Ȓ˼vk/wT24 J?屍p-蕥*2ӝ/O<>뮃Z!"qnoocd1Q W7$($dr|F "걥_8 /_G}D@1d28:UNv.ϓ K՝#_lD#3!@b#t6_~ _u} 5ό"iK$E.EL$$$(  ((;whFg )5 mUI8Iݻyh{XkLѽYLSmYR,,Ðpir.BflVuIBID,ߧٖy- %Lyoܸ41!"2H:e'L,^?ڭ,=lo@+++oYWܻ&ʽ=>J)G}S눈'NdGϟ?9 _-<(ea6BYhz;cO^pJ)3I" nEKs#F'E6"i땓;举Vɽ#0Al!ns뭸 > "%, ◾0 }2ggٖA08q I:HQFQP JL쬘FEUgЉ:6 *kaDHD߶i%e^4KEҦWy4(&2M?-կ~ƍ a2S(,aLMgE$пMkw2|ap$=y^e?^w{1??'><#st) RJ'Y~#(Vs)8@U;,O5/am %o{ů!US3d cqN&0 (%aM%;6ʂ$͸mTj.qm@to9Uɯ0|S 8Z.2l U||YfVc֗+0ʯ,>2;Xr(ec ("QdRI{-/pN89~zG{%Zu}G?я~z~ oxOO_|衇$N*;k"L'c'% O ߈hEk zi>q *DdYБd,)Z])-Z>1tzw48*r܌67ߌ?OQ4:ݷ;*y{aDlIۇIDhUQt 0K>`YE$"\Dx`!ݺmWСۋhTԗRf!W+nf7> D<#`ʉcz<БVXov&& !K+,h]'qwG`J"ҮeXL8{.FIHT M%AZ^jC=|ѳg,Їӟ>syp%c% y?"<",v8}{μ\>ڻ#QuKc Gpܲ?Ã7Z9W\)mU2(e*pbk{$rN D8RETÈĵAQ_bGu%oGQ~?g(CF-MN-Bu߸q#orI1"e>EvA$Aa]Jn vWJR$("/_e~8}"ƒ)Ssd2˯6SZ^+n\ 曓F[Z@<Tť?M347A'w֤Z>Yc'l,<J)aeR#`^ 3pqD)Po";ǖJJ#G JӞ]48'"U#2 !"ql" ;; OLtzBnShv `۠i8e. ʂ&D>&˹0YzuZ=yGN BZ‚qlec?""dן BaٚAB!$AF_ oǍmQɯ "& "gUD58XXD8ݢV !T_b|X,pMBs3հ^6,QiNJ4boܸM|-)?'FFymL"B4N}7aL?Dxv>NBrF"BsjtzG7/pNgb4꺥y+Z v@R8/W*wsOV9k'qRDR:6t/:6="#,U޲(4ϋjjdkK)~򓟴<3gT qT>S.jajfFK$J|N2(PN6"TjvPJ)혍 T(ɵ BUUIM؝^ETlm0$) 7I>>\.7۩Y_ʯ,|/p8+EҀRC=tF Л#y'p`ú*障 (ciy>"L3l1J#ƘI!%Fe]o^ {ڟYl u9t|Tsqr"ء oo} jœhB@@^+Je~1wj{C'j`a.GYIk YZvTdI-FL[qv72FYTLL=SGCE908QbRw!I zǷѶ(2DH_%ۏcD^_0YT^̩Ůռpnm鲜ab#pmȖM>+Tܣ, fJ$Sۍ=7C_5 "xĉ0#IMe1ሰc6Z)կ0TG{g5>"DDn$tGo#7 ߒGhqE^BA "B7prx=;9؊8"/-Wwc[ͺ&i`QUA@HG\V& KlgE}ɘRվ`r|+cieeA"Ed!,K̶dL5 4矻3Weh? M6(e0fYH*s<Ї -f՛a#$p]/}/' GVXo=/RD82rR~W`En蹾"4y$i*BxdfEl"r(0(ogbQT˅U2dtJ[3H58 4߉: r+ZjqĠpw #mn6 K|SqQt㻎ij4 R \w$LYFeg9ZRB\Ҋ` 2 2Kq= h$ϲq8F w;qjV+|C-ؾ^xM-kf<CP59΋mY"X2H_#  oc_PQL1AnT|kD<'qM]?b=K6iD""K.rXZg`)첣On~?؞vUiYe GYCbiY(!0uu, eݰXK~0x~ 3v:nmQץC%čC;~aQN ͢ ,Ҵ^L77kTL"t,24ZrrB8o;/в}\n_OҒ "d߇%dh^؞95@@g~?_Q:ZGO?MS/L |QJiN-G78/ڦ>h8DjPCMo&ah "n8'4YnjoٸǯotD8 |"Cj#Ys6~DXO@QJ/\L)?wjRj(H eD|o5j9>yR@$IFl0(ahiqK v~Smbw&noSpmuQDyJ[om<'t]YZ-P(gPJzQ_9Rj"mbM ׷ K.ݸ6biFH#Dn~xs_ʯ}ckN6}/zzn`K8Bf󽱍v|-3QrZc6R z2ۃ >9z3闟?G,w}`ow&xcRnvv|lz=V ?#Z\gmYh#"Z$feF-y?7Ak$klB_EBEA]OчVbZM]1/,_dIP,opg7>7aAS a/Mˎl dkÏjy#5~2'S r!!IYDgB](d尌*'58[s9MQJLaƀ:=RJ2V,y Tr5xli#7/|fCd=j(ZQPX(wG>;t "LTJ,"7dVA?Ep}tww׭])׿݉67O!8Ƃ^\("=G@D`#" bY 8V63JM)5K9:7M2xJ) bQQ4'/p?9]I 4Uҭ"Wj|(E8GB:58( $q<^"4yQॡMms{bz] EAPչw.XU;fR bs¡vysȽn~J{W jѽ=u;PhGHf>e5NXmz j_P}eȬi04ɢ{8`O+n/77i>,o'5 uJ)} C9]Dr4>ЌeW٧ϴRJ^LҾZXk vxID˂K(eƏ?]-ς yYPD(IP+T9!L5z?od( Q6dЋ[ݹ_<094@1Cxl# BtL5~) BMۇp_EfME!K7yG~Hs]-llTNoO~&IP*h׊6p,>mFiQ,EQDUOq"V*m)z0iQFoٸOR4gھN "K<4u! -l)"9V7òhQ\:^={BKm0ki@ {!tWBj#DD]9('v&3"& YaKXBEF5g>6ܡu9Ǣ$=߉%7d_Go޸ry?t{o=׽ ؀b<{!Q w/QERIN|)2hGqw;bOK.T qXylB|W,%F3NyPsǗ>um( "6\|׃|; No"bO;Ε[OZ6܂!lo"%5[#b.n2rTIq]Gyg(GdΤA$y[ux~_p * T ugTn84 motvo?Fw.|{&CH#9"̮%9~/xZ!"iE[oN H8bu ")ZX)"@˂HPQDMôUe= M]"i! Lkv'ʛjt/&lI!J2;fl,s쯽CZee#/}#Z :$߿QC_61`T>TH5U0D;ʎFQ1uM6ʹvҼ |R <[ith,ۣ"xtqf#e0Ԃ(plNxK(b3H Ւ0@5FW?.C`.)q([ 8+g}8p,2_]R; h= r[ lQo4hg 3N_`rTIya9 Sf"2gR],",(~@ؽ*K\R쪲,2o' %+9jyTQEMo`(6r/~ؽ?pX,Bq(3COx~ziӸבkI9T$*lJ%X[ÝmwrjU_;tz{9;[m"PJK%LQ%pדnywr"bZoO! R'5b8P۽=HλJO<DgkEBxYxl{=Pn15⩍q;gQDZX;}Ǻ&3D*9/5ID(O f#vp~EA$Bx3!^3l0gty{a=BI5Լmcp8?Nx;ID;c){ր@1gp/]zsw܂S0n[ҍνCBe8AUܲ  `ҧ_b9eԪn}`0sv1(^:=@,'ɸ"a0p BD4Ԃ)aD扇 <)ǁ'G5^2G<4GP _>TK5R 1^9ݷ7W0""Xx}$Y(3{AA0i#S?,M7JVVq=̖ W,?rbCY*;{u03 @=WJf:e-׷NdY՚-HN)t\Μ/~q,bg+!@R]y^3jؾ[DZN-{P.L jqj"/6z䰅0PJK%pEi O>(eB&EN4rEjwH0*ZtuW' g*ZXN9M`cN ",/'PC{:[YBe (UY}Cp+NOTozmL BB0Cm0 %โF3eJjRʊ5Vr+$ NQj\XS%0|^D-o $Ȳ1A7/|{w#밲t̙6B8t[ޓloFT5jkNZ ; IL0 a12҃# P@z#VA}Ib-mcz TZyVRU]Q ,@,sBAp]P:o>W_su&vAg(KJ=~ ήW} "8 !ٽ sEDߧ&0AP,"?͡Tv,n GfyrB9!в!B/pOdP ""&TʂCl!<j^osgq~3+eq"6B0 4^׃R e Lkﭷ߇bqqs؅oYКD% x^D86œZ܏G8LwbrF !&;h55( %d[ª %c5;tTzW˹eYPzow\ۣ7`~"/Kl{s+ufzbuaϞf+tΰΤcDX,"qsJz!yNPD<{^Q˜QHB 1|>qw-;1ͯV̳Tpw~ZV S%5̶l6)a3vikr"Ta t0)4@}Ge$Lg&sal*+_ILw}~ \3nӥ%؀xz,8C^9UW~Mozӽ;}>vmwyLG gmq JonN-C\rd߲~m`ceHD<P`;_V$b{V|uiZ QN i\&7K4zuW'꽭88W[oZ33u"ht:W~hYP*vdU%ԊĚr ^4;i#A@i$00-¡p@UE1E g!\9QvF˅pЧ\-@Ģ^@g0dxIDA#8 Jcil]Y58i CZFb+!ps"B "CpMuE0HX"OX s尴2Zmn)hne,kyMo>}/(ٖ&u`%\<ÀZUnCTU1aDl/;j=G6 `4 $ gܤz?6?x?#`%'߽ @KKq_u}-oʡbD78S?G?wFLZ@ 'k#D'pw\˅jaلr9;:[[4&X.U9#mۨ54`?"m焮dGE)`,r~}"Fkt zowmCjPo%n,kn9R]H;fcZ 8O;<e鶗N3E!%~9hÈDqH" ,DKYFIRAB&ҤT⋶- h YmI{o7[/d[6)\ZWpJVɧa-mł0oFEDET@DUE^ fv̺[[ԶJk5n;~Dj;QEgHt.CV"N"C_ňg~gN6=j__4ǔeE%[9QC.u^$Fyj5dGJjō( f}/pcdRA\Y}o)-pXL= BscNE v|K{"k=|a+ܸ! 0_)!BXZv{44"^/q9ތ7V4AAׁێͩppTHdF6,b~8e.% ˅a;Y-^z2?͝q>*_{"\ QvA-ϢdOuYT۴PzW,2$(!7BɓҠ2D x^,~Vzbo\TTe7 Zn='AF|O+W>0NW`yIDeqlCzefʾ.|JPDC"_!"t)6;6nw/4nay)<_Yl[dvrN>ON8qUsy00"P"(Ձr dI^ q}=ANmi(^iώ&wtEE~|GcE^++4WշͿI&U !AXR,bj C!q^o}ڝY!sYm?4qN7Mc`eۣ?<7+4ڷ{ CIљ0 $7=3~!lTrY.!<ZJq,W{f+dvV ze`w X[jni+_=woZ&x;DV֏;fg}qgj5 '#輳8rN}9=DҍΓct߯a$ g|mr@pcC~?B'W,5Z'ʕ\6x 2Bo&s0pla"4B|Gݞ>oKze?mcET'}NK<~zNwLoO6T;ށ<$5Mur J_ňpzEX"b x@@$ D K)elԊٖV9xC\PK88"d|4( ]P Zs*B'e#ҵ 8z@[ӣ2F !Ge=qre"BD8}/^ܿ'@H=W+8ب*imzltd}wZIz{jax(F++FJn(&X0 ޹;ۋYd^waQDYΦsxxqh47-Ù3\qxsnL5:Ge] k ,/c=]2$AoK(ju?!dhv}2n UEn5 8#"9TٙJ{O<Ƿ ze*-`c#Ar<~+ +r}7@snr/an4B;;;?~{o-4 P"B7]` ~h#:Z@mvɝ3[ 3<nXs~yqaĕE-{4Ev" 9"F^?2Z:1i˅(KO^kaVE>J3=y776@?Fp4\a "Dߓ,o{R9nR~ BLM3YYU6 %p~rЕ<+\]~GOw,hm8ۣ]SšO<Ƿim ,qע~u|xN '= 4c "xPΝ~s'csaC,GԛyV'+FrvKw~۟_Gֳ;ckAlPRb˶B!'MD,].)KrjNշ۲1ڂ%8v i%W :JisR:V>q~i8x6|InM4*Fm[E),qx׻z zx=e- U۝LiF7tm Ck ]ZίNM_}E/n yr糆WZWWF!/כ^}_}=kRIIx_Fk&WvG5E&[[qϾV% N\mcS/\}ߎ n/+r/< OO|{~ԩSƵ0ը<_sqrg[&q6JVHE $dd@#;sc֙ g_F,"ygvvrf{G3#$=g~}5izpG>JuZ[,LBckBdE!Qv d7Q% t: s >#M)I("q<Aa<)8@ ID1RJux]+Pq{Dp ǃUi0pє\٠0>b`2 ZY!$C a~(+±@r0"sīODO镩,"*b0'x>)ҥ%TBi}>bA+yTdk8*GvC2 ǣ>p J+B&*@ssji_+@7L?4}Uo w?p(&tD#$@D"d!̃snu˭a`D#N' 2QqJ Ev].< (U@diɢnX> n4y9HL[Oq`1I]nL6D#1ȶg6H090op@%04!ꂲ2d2!b hu25'Sں< qSЅ/w7;_|Q!ۿ#XБ#t!eі#nP$au+FQ!Tiou[~^Odk4=B2I& >!adׇʡڱDEhz_iQ #D~@ɱ? 8a̫ȉb}ŦGwNٌ3b4JTv^sdrAi)x< [1T(:ЯF#Tƥ$(` ٟ8HDcQ#~0 `CKq #jѱ^IA8LV`24W-La $D,"AB(S'x2&0ۋf(+C{AGMQ=%$غȿLVL^$Z !P(`BOޒdS@*%GlRH{ExhF *+ezdx1@@l9  Od6#HUNԄ P _Ēf3jlSwR;ۋy!"2 Dʴo`(y]"nAbq`^e0`ti >\V0&Iz7VWz%:<"]y'2q̘A= Owa!˲A0L !$˕Je*X$J3Fnh#z=rqͮ_iksɖGbDTΊӁ}(6.bRF^Ǭ)I I!`Y‰Le2PȩdTt1eBN6wX,DZEc)*3.S;bG1HSVHVBR::QN0U."H A*WJ%Ǥ Rq(ˡ6UHZhvZdȾk"X">OFW{DL8,rJXbЇ"doPBe`xP@r@00VrMxF|4ɧ1S>H? }W k*1^BVU%[Zhc0ZѮ.P*sۘBfs)4'RiBd!q*ee5(Bh+ ߾.J%b%Zw8^0ǂ,Z,I$B,Q(U I?pv~R!1zo5Ǣ)\G'8u8W] &l:QL?sﰍcl 'X|\Ƙ~F8BkȢDDb{zlR6(_$NऌlVaIDZDt0aC1 ͦQ5qXP2&˙Q@dLd CKF< CâXbz*}l8 /7Xx'ݘ0D*2 v|w;r>R"D\^X8Hd /BPXw@H玿rˑ}aNvgxʈG4őxhlp}"]9JhBbNNK9 sP\ &B@L`* l <Ė1E~ZXqhD_nՠ6qU,&"r8 YzYZn>~dѠlhḁ 8@:CSocԄ VP"Ak{#G:DyףdT 8ҏ bH wB.N$pz b#TH8sGyC:ΘZ,`2A$osz*R˟"{-R)ZE)T \B߻̩3Od uH'eo)FD< l&@K$A8 Id4A4;3PBHʴC[bTP`{:G1Df= *7v85˜\Z$4vO#AfYQ 8±@yT 2=HC6J(dfK'LYoԿ"̈P*! %af!S6ɨ.AY~\qRydi%!U8p(U„rJJRy<4s25Ԗ`GX[HD!A9ͰlYjpt4$sp$.$d#n@qYupujp<ͽOhl 2p,eCR{$W-<}ru4${mGu\F'D25Bv 9"eL :EyRH#~ҘHW.-@DRij% -Q+tĤY8f3gR)~( ^,R !+K IrVɵ x$iBCѪS0nΐ KeF͎j;8(5b0C $HiqJ5Bɘp% ̛ wLVR r]HG]Ч۽Dn#AQH9 >FH8 ` f~DR= 2BYF%`YɆu.x͆p}n>_֦aeZޮ q_nVch0Ä! 냢"AʠF!A1vlIsAz+󼬇V!ɉP8?d~ ^5 ۚ\,_-h:1ǁQS7-NExYC 0dXjYi ^8C )`[$Z*S#\w &Sq4X}!w<%+B)9l`ͅL؟JE!)KloYSLQ бþ}$(?% c(h'߮6Q3 d ׇe2&=Lr9c[p+F1v*0P=xC, P*mP65H41Fٳ$&u1jiɬctZ%TZ ,FR%('+^/H$Ĺ4DZ6rav>FH~ II"P(P!8>GDfvI?bm""jGL8]s鍯\:pMMDmT5P62 {,F|& OFSBB(Ƅ5!d0ƙD*%t ch#~ +0Z$S'Tj@F '^J!R"M31T 2|u:@9 h5Lk'~19E"XNi"!cd2N0 L̝m@[4 E qxB#!4cu@j!#[,(6BZcLHa@At֌xС!ۍZPUp׮c~Jb8;X WZ|!;/|"g z,xR *:y\ID>-1Ʊxe$idjlfYR'r"@.Q濟caJ mgswJhq({x beR (7˅=Z-+H6ncˆRZJ<>!`&aAZpP#ġd@QHMݎ zꩧfΗ W# 0 '>[;z+S~q_p|*_B5?c-<\B"ƅ)˥ [?LgS 0,IZs#*($cߟ)MiK4 Áf*acڅ?4 Q>G#(y|tUi3rzP@43XR @aW~Sdjb3eeamVw_ۢ-/V j#$|{ 'OL+ss|J !22O&Y[1A.tCfy)KX} :昍WRQSDuDca^5բDTmՕsEn7` `kDq6ӨgĀt4bVF c|}ۑ]WNe !B`_S&WnթMu V9HbT$s2UH+jNBؽy] X!GKYmKþX1N'X7S+r,pX"Pt:&f?&a8~A!hjg q!D:=Nd# RVKFnxPBRc8v /\HE  ٫}qݍ#~^N(/GvCeO=]d-ؽ0Sϼ:zqBp#yk841Dv}-gpƁ@G"87nV +$e Hl#HՅɖ()7ʵm"$0mVoތˁ$(pԃcƜ$^ϰ*egd"(+*cfC==H"P( ˂LJ}c aLDqB -JKY_L`єm`CQpF D!UԯLEF{{qq1`*\4% *7w2JD}`R)R*-~ …czEML!IYH,fIb3e4Y C!lFr2'8ѷml$0]R)z1cS~/ţ &;YZ}\[M@̄@:;qi)Bѫz=$υ3TC; " шJ\׶e-h9ֻgu+F^$8&D Bm-ÇSG2V!FVW#/nD/ŜLmڗؽ+hzRjTeb?@Ujccxm]fDldTctu%`I? 2[Zk,7Hm8MIzm먒 `HkGb&%˰|F,㮮h 8GDlKd2 5cF.A;7sP.`VbH<*|6BR# cp@(-gh3 jK3G6AȦ?+3yMQñ0'J7ǎaz7`{ޚR,XB--f v;TZ??>O'[[ .OMB9xԌF$cRpKnpt: ٚĀBzTOM܎ ?h#laFIQۇ l!DrRj9c"02 #-.F$!TdKP\Bà!wK9T9le2&B2)'e#D XXjz)+2O|@FlRmj500E7lIQ`2Ar9%h,w<-g"x,P(c?!d0hHu{j#ap6لG_p$ϯ&Ӄˇ$d.¨&E[&JF]n3Uz_d>qrFKD}CP ebjna`X ` evk55V\UL&(-{s'V^pdjNN$k0\ЀkOٌ@Uiq5 (] G+.DǙz)H\? #^a0v,2=㈥Jx$;Rӆa((+C1D@ ]VH`} eGŌ%[VH`CQEc,e x׫Z#hҘE>w[7 lēx2iNr>u M={eErb&0b&R4CS,0̉i"ϒc| TT I|=6B܃z1Ơ4BȗUh,ҏwKIX12I,jk_yŚ`0 UhKD kj8ISܛr;qrFyޠr|> PSlm% goM jn>N҂kjEĉh.\]q- v8FJBF?1bcQw7xCmv} j 2LZуRkL0,ahvZ<QP<'Sm2r*ҕz: EBh<5jD z^`( p . MRVWluRdXFr?_:_&]Me4^e޿W"Bk Rj8O: b1()9؜2#Dei9z54(tvQW#hdjnCTX aj(+C۝nZ[1)I2;;S4'Gq8OAeai8S Y2 dt s2nŲuwc'!*WD+$+XX V0,JzK8x^^~@)+ i4P@_`YF*a~{+Z# &"hO6*mޝ͟a#  $^O=qShP#%"B-D CE$'TQ,J`I$]])vϒ~aʔԝKa$L%Í7R BJe{ :ƘB+m<;7O3lcuJSgKь qh w,{ܱe-5arbHQ ~MqC5 rgZ^JDקDZDdU 9cJK-~u˦xǍCqĝ;V㛑O↮f4qZatӁ74K U AnO$tx!BȢDCryaJJa2 --9HYnU%Y!>_p"lv'T*8-hD?ů@n7\LŲ"hBwƍ(NorRkr5BH0BhD# >x*Պ| D|- fooKD\N:JgBIҤ)JO]/&D$Uh B$F4aFbCw[%azK)BǁҨ4W9 騦3+ Ɇ~$!9m {JL1c[Z0 jL8KK.B¤L Lej r2#*9mt:H&!VL2رq)& Ғ~lm&SjeawofOpTEv8xLD{E;tJLUjoJ$ȣb*5PT3_;I9Ch m*ҕxv~+6t}fmNҠakcQR &hLRVNQ4W#h@.;S!vA \24Rۥk,r c޿?2ZdJ7Lhƽ5NduNSioo)m2G^o*(Y>W QՖ1-}O~k70 RMMvH>f zܙ`BرԷ|K *<ލ@駅ѬY1aMIjo}}l6@uHMyt))$BR:!˪cUyPF xT !Dr<BC[qwf w ?{ Z0>S^\ Ɏ[E%%pfTBR*b:#8|o+VP 2j:!`NJ/+=m֭djW\ ~rRL)RAooj_}[*/GC3t=_ȭ:hzR #XǕO%FW{=>?H$H0\)c?g<9X9E0YNPSir֩-8F򇒆)$,Ƭ-͐@s3 Qqf;pz TWe5e $?A a\_uu0469)zW#%LS4c٦:]w9xBls|4]-,' .|Qyo&DL\Ne(b˃ٳQe%zcp:!oA`}޽ˎ땔Qt5g7Z 3^-„ .]h[ټY3eg4 |\^/|oT`rBr.RVFS41TCG҂G:iB!LsPG"ۋR0W^n"`YoަI7iD2v.m)EMp onm>\DXm^)lLr]__{5WwIZ  P_wkZfCMM B@BBƲ}&W^RK ph:A\ᜟ&j5zZ"/g_[.s 2,.̅ a現?^n%\ʧrѓN\@} }7" 9M F4 Fm݊w:;qI #,C07cS"$W:)]([n ;f vp=|$-}_ssՅ6En'RS m91))ZAg{zt([~ B4i-~:ߏs {}åD*鄁r'qЛo SSG$KPf@ˌd]tM$RG.QM%fPBHRlIƍC}}rAk+TT Yub W6O T*G?ylFIJ0 N 144 /YBeE;1Q{LXn,qB־< ͻ81Vdg[KSZ]pn,#[' ~szjg瞣ozec!W)!R/RVN!I#2t焖!B`6`/\2V섷Ӧ{索 ?cQE1_qغB]NiȦ@5ٜ޸:[o rH4¼A _*EK( 2cέ--OqIq0ƵH*Eyr(.L`dRj>jjB?IA0mZO xj5KԴG}մieFCCV+h4؈3CKB;flLf\zj]EBÞ˲2DP=^-,[F]ye;#/I=1Kg9I8Nw5f1N|J)eaBE uǩєACp]5P[ᅬ-;etydexR̜yb#D#ܲE8\dP|BMcx]A"7 &I(aw7߅?PSӦQ8I/K ܜJkI}!)cFGM衇JryAGOF,h{sh9Q~=%Kۓx<ǩ UN(]]xF.hno`en{%#B.+؉B`̘AgorU 'g lꛆR ho[oEz !rrPѣmތag<?q"=[t9sЌv_|!ʏ3p V !PVZtY:WA(/GwA]}55zMI>n :hY^!5*"N|64ݻqg簞2k\k=8Zg\f3('\$!*7R2ʕԒ%d$sWG>JكyO$ ^NEƝ|7O&G{?}T*˘G_TxsAOMpBA$mӮ&4grDQqPWr KKJBp>zB! 9a6M ޚX2 F^K=t}71 a|ce᪫N(T?Ɓ9ImTef*?Q?(#8. *O4 Q8|GP};O<~G3̩жЯ~E?? ӧS'̧=s@XA}bTQMܔ(-HؓZhէS@ !:IBơݻq$"qNL pT:ľs^ihw݂ ~9a6C}qQbP^~^( jtčߒ2]UeFZ~5TV:V+`t9gfHcc9jX\vulEɓǩZ ѣ.D$[Ʉ_Gj MMMv/2׃?5eԼ@1q"zaj3k N|UVm޼y&iŊ<ϟh :%`6^r9%@M0rOf0)E*g~hov LH:3GCW^-/A4oh/˷݋L<(=}zjW;X 551D"EEp\y%N$ ?œdw*ʯ{XxzG`uN%xO$cfҒ6UOxDW1˞pH&Lg ԙQQF555z3gwmX^|'475A['Nmw[ZN&Hr ôiӦMv,,,d?݇|H&v]Ӊ`۶m6 At֮][ZZZ__ܼjժ{W7H%wǍFhv\PHPh4t6qgqjF`H~mmmVU܁=z`08m۶;vL(cXwwwoooz;)Hhv\,.].׶mU*w>r4>6mڄҧ {<'Hofs ?zEN=cÇ+a4S{D6mnX ۵k7{8Hk4UՊ3m޼l6_fssFpꌭTh O9%KTWW,K<_0ޘ1cnE3G*//Z+Wӟdt:O?}Z,+***+++**9s\{˖-;wn0$򎎎iӦ,_?f۲e XO0eʔ{BѼD[WWlٲolƍ4~gD;3z{g̘1({_y .['O|x<1nmm5jԭ:s{G<wF+++l61ްa%\rm]z饓'Ov: B49sҥK.]:{h41~',YrwyK.bG}bŊΙ3g?l6ێ;0ƉDK/]d?|q| &b˗WWWK$gtࢋ.2 ntL_nw0,))BC AX|c=F䍍s|7۷dq!x<&,GD ƍD`$YjUB 7pv!*+++;z(in>tPBHc֬Y+~<<`^x9F1;~+z.o9sL^lٓO>IyWcxdEEEBO|`0fqIT*–-[͛R{͛  ڵk̙9ǰ8>;;;G}:/|DlڴiܹEQ5w/w^Zm27o<}tq?3{lr͛Gu]~j,w}o?LhFW}ݻwtk׮%ه0 ?ꫯ>&wygɒ%98 ```0F8 644ѣ>!e͚5}GGի \.BP(ȿz`^qb5A|s@ `4rPL&2 w H]w^lc1`͚5Lovhoo+} i?3Zwtҳeeeh---moogpt??|7c90d@ X__1O̙\!af͚wr}"cՑ)bY~ԨQGyw]xx 0jzѢE{:#G;S5nW_} 7nF ]f6~i4ݲe֭[\s͊+Nƍ= ׯ_p!9{ƍNß|I4---EYp EV"7ndΝm۶m޼9}g 6dC.\`Z?۷K7,)H6˲W^y?_/6"}'x;ܹsƍB+V+x MZ[[W_}5yqu׽۶m۶m[o|rxWw/߿kiiQ*ݻaÆ۷袋`ҥ|gϞ^zoڴi`Qr# ۶mۺuG}h>Cb<=8m6mڶm… ?#z{ 455yǓkijjZfMMMn'Lׯ_jժf ì]]f L5kyfm͚5~͚5ƍ;Mx:oz߾}wjA@dYYYJh4&%%l+W{* @iiiNN,Y^^766pj5ӧO|>_G<uJ0gǏK,Lp\6h,X]yJUUUpÇ=HxҺuܹ%I]]<m;BVrrr]]]DDtMMMw0  X,Rn2 hܸqcCCÙSSSW^-Hlɇ'͛7J%r\& fgpjuBB|*+ Z-㽽yyyA {zv'IRѨT2 zzz^|aÆ"l"ȴ /^,(JNNvlE1}y"ƍ 4/,,lG~:ɜ"˭Vkvv661mxxX*I<<K /Sb v;̴SRRRrss1 KJJMEr6 3}|Џ8 ‹/|8͛/]{??aNh"@PZZ8F!jd2Y|||FFInwccH$hL&jRvGGG].Jw~;;;VdUWWL&QYY ˻nxQommAAAjjz{{K$Bt:t`0?(#ȯCih]v PVZcx֬_$L$Y,T*rѣGp_|yyyyjaaa``3g> endobj 368 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/files/2014/10/nowcast_errors.png) >> endobj 369 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 370 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 369 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 57370>> stream xw|TUMK'@Xa]AW,UUYVqU)* NM ғ7 äMIf&?grʹo޼=K!@  i@$MC !@ 4MÒd@ e(b Ж!!@ 4#$Bx@hӐQ@ ih(H"$H"ۦL@ *dh@ m6 0ӧOxOMMJĎ;B"B+ŃB,rɾ} DV0SO5jvZ4iV>AaB믁-{.,???%%%'''++Ch׮E]a%%%dgg'$$ݻcǎ]tsEQ=Z\\թS'Ç 222t颪jII$I{Kv{nn'233NGnl6i{q:YYY3\4܈v'$$3gl6[nni}u}.W^qqqx}3;޽{].,EQÇ^xطuرc:t 񎕕{ .oxq_e/pB!C27>?0%%%++{ڷo?iҤ̜9sȑv_~aÆ;55uʔ)oڵѣG-[@tݻw'&&nذG2eag޶m=o~홙?k6| 6̚5뮻믟x≑#GVTT 8:u(]t :}8mf͚k׮ݵk$I۶m۰aCJJ}t:zݽ{ovm޼y]t駟|/㙙˖- 'Nis\bE||Сݻc͇ '|g7o~3g\wu]v^w7n\^n^zftݻ/_ɓ_a(8n-wuc1ѣG9{'x4%K\q)2q[no͚5izΝ;YCYVVt:k;aNsϞ=={uݻ( 6o_^^nfiiiO8a۷o^{Ϳ}iiiFFt'N? ƍßo?4o'čK.>}:{͘14޽{oqРA뀯LKKK-֭6ln7or\YYYeee__g~ǿqSSS^+Cm@,!X`پ;V'v  U /oر~Q%⥗^o߾aÆ%''w !w뭷^z饃4iѣi/7o^\\˲eee,رcĈ<ӣGxӧv{NBvvԩS?QFM>=>>[nK 2iҤk6Sǟ322qٲeˮ]ƌ >xX2//߿?>>7xA:qQFmٲ%77n]O:(JNN?(.kn{РAW\qŌ3.ݻw?X+2''S{V1bĢE|UW͘1k׮$˪!DƢ6 )) ΂Yf[y^Qrl61n(jO>/2{lF?άY֮]]RRҧOf 4AdYƟUUej^tEؽ{?;>[pSO=/̚5kҥ^{m?{zjZM6sLE9w}7997b?eJKKׯ_?s{j9^_XA~s^y啦iv!~cǎ}7\sݻV_UՊ?$SSSlrO?tĈǎH"#+cx!BhǎVھ}ӫW޹sg#$a8ˑ#GxԨQoۡC.zLȑ#Ǐx,Yd̙(yÇg&++kŊO?tyyyJJ7`q>x⬬?3g0ڵkk?W_%g={ܹs?Ν;7++ u[u놝}&''Gss=np˗)]tђ%K7MsѢE'Odꫯ;v,0+W| EQ:u!C,XO̟?ܸq'O$ɓO>dee  ,СwQZZzW,O\y0x`-P5tPKJJ~+I9C$Xj>>/p6믟8qiJJJ.;wo>J!P@4"馛馛p&))iYYY0xo8C Bme˖=3o t]X,GIIIeee6m׿'G NUis=}" #VK<뭷N>8r{t]=eYѣG;6Z Y99yʔ)񊢬^zȐ!}n6m^}_0SPPsNRCkq&jJJ<i2dH׮]I*Y B 0Ç8zh<ZrmvmoībDQA ԁGZ$B iʕ+?sǓ=cƌ>}|6l禛nۗ.[lӦM4M8svj<_t:8n&Mׯڵk'M4iҤAաό3.2xWn݊⋠GHhM[lҥ 9sΝkZ 6lؐ9v澌%Vcǎ׺rׯ1bBox$<%..nĈ-5nLQTǎ;vX/G-ڵOܥSNMәշ!h2"4ò$I|#MOw8ݻw?y$'O}5'ZBnw3cX5rݦinݺ~߿?p w>q#&@ X0\rG}Eq޼yz~{۷/EQ˗/L"5#GtZ 4gΜx<6YRobXMiՇ5Mc٪i!k֬߿xb&3,\0ҊD#1ȍEQ5fQ@ m2pD 6 1V D!$=3ւNbxh@TU4_8: !@ 44-Z%&4!%Ν;MӌMleTMPbAI"];"lw': |Ѡy]lo>J 9ƥ_yPB .h[+m 5WrL7Y Hn:CbDN9k,<}5!2ʥ^$͜934BM$YfhЄȖx5YTuŧxBߞa]FQ(1ѠI$>A 4.{w Db 6TM 0a7>lr޼yaD&DFTuA1BY-Z GD& d!SPa&2M6L=ĽOh?d^24J .@ 9L'!lόfʔ)dyzB CĉU+B CÇѪUm<BBBS4MѡBجZZ S+DQ@" ,pEQ>2 C^[!J !TVV5 p6GhlV- 1U]6ny0"!QuY y*7o^jEhaOhVMZR* yM"(Ei> &sa({(mu}۶mv[xo,D*MPߏWhkk%7QjU/Lq}"(i ="Fy,D!gFqe7㖝j Vys,EG!ߙGbA%q j4MoٲE4m4E* ,"QO$^Qu( T " .Əd(Ҫ MbcxL(Q#k]Ru%JL4h IC׼(jwqǺunݺu۷.DG-&G(Fۦ?>0ag=ˆڤn'S<2 3M"'gP!- /^;hD& l5M9rpܹrbŊԶ4E* yJP4 yD^&~j ^t5o;R34-`38*Ϡtg aHL4hެ_zc㴭 IIQ,QUy8B}~];"9T]Fc;ֻfUG7T_yO(N9`TBMW$k4,D4W^}]wMKiY5h۶mÃM9n[GbΊkى Q֐5sE)6ꍺsΈZ 1°(-MQԄ &O_ׯ_~4x4S]420Fh't閝6!pz~Ohq<=+}2(0Pv~֨[5]эKk/Yz3~eY^bE{MtaXd@~~io㸷~[ YFiqUEo*=+80 Sx M!(`ҥ ,lIII Զѣ999 !k׮W_}!0L>}vޝxʍ t" #f:hC-ٹM!77SNM<p2|aqt[ɇf{\Ggj2C{v)TѺw4M3//c"ai(\)۵km۶^zu9///(q .<|!C$I(bU@t:l};:2 HLZGaG?hFp/nbp͇K.LKQ[G2iF`$ !FlP!%8}u(ҥK.\taiY-:}+/_ aUhyu[)&N24.pڴi4݀RWaQE~K!di 5S,rI[2P/YG_nTՔ V᝾D[Jm; 0yF+P|gۑCfhߨLxV uNz,wѢEGӴn޽]vopu麾o߾ ._ر#=޿޽WXQ[|*xh4#B!SF@$Ujj-4}|R5tСC}dY4 4 WIf0[bE3TЃBqeMGqɚD&+4CL PIxwIlFH1fh@xA?`Su0:x*oP# "'|Q&, -hX,@t#ڻw; #+ CR-/epQ``Y beMYAyy!o(.MW?nxV)+flhzȋ7n8, ,Xǎx;hYQr"O7{ɯ<BB(*>>}2a?4iK5/PTtOגj2n6U{0 @ywݏ6.ЂqÄB]WϹhή,Xǎ5&]'q͛76&`EbNTfBAjz@LTk*sD!~tCQe~`K@0<~6fn_e{fT5Df۽3Y-M= ` 1enIIkW8~GT #$CM-B>$ED}YRGl従[<ӊ !oA K.6ouG}LdaAS@N&^I%?Eb"5z<+24SwR2#lI2UR|P+6 ћ,QU_~|_~y||],(REIJx. k>DALN6qTtBMW3'g !C37}_FFeb"Yu ߷?L薝<+3OY9w˵&z$X--@h/IX'N V(:ë[e6?{:Q*ԼG_~DUGv򜈇z*Fԡ*'Ÿ:fPxdUKyP@,WGN$-##o _G1Bt6^)ȳ6GkU]*Ч1b5\lٲ_7|_|Eb>+RCxxu5DGZ {Dh E9gjU|yE[*BQѣjV:=B@R:L ?~K6NgBZb*;- :ˍ Ey:"34fs _[whq˻dQ{$Y%#GB& d5#leBǏ#09/=0D!%Ko˖-kJHQTV UgMR5%#c!E.Xj0zA#I?USQoGqYn}ůPл7Ur:lV1?7deG9pdG%lmoR .((*pt!{eWmn{Wàz;O[mH몠 c404Kn76aMӬ8:04ɆQs֨X?ɗU/HRX+ut;vH,?{p6#LqSowwE+J( 2&HO&TR n7X`8*6fjK7M3/// ΔEBr!ը*MQ).tڄ!5Oqĥ4 V=U/p, a\zySPI ae"J]"*0T=JN;]V*dT<'bg1B WShz)Ӟq,U= ԅy]6c$UyYYpX=1KS|A>u.5bTE@ uBNoA7pm!Թ$] l6l!%Dp*a":[HJ M }#lD3h<\iU`O.Pp !T4o⨬zl)ayV"LԴ@0xBWgi7h2DJ]IT,g/sL=8N r VBz3 çxq_Wel$j6nW5{Oendi (g:w>g) P, úZU=j@Z=³2n7T(-z,ᇍET[O:tDZ)S;${ 5 tS:*ѷ)ڄ!,sJ`-hhT$^=4MEl)ayV;_#TuEӕcuG74PuC-,hV${{pRU$8:upE \!T)&DzCcO5{e>X,Uu➢# U=xӄ !6ؤ `U]> h^E8q!Ye- Y_CQ/0vzwOH)#Oac/<Ы\ eiĪ!4M֭;tPZVV?գ֢"҅e4]Y޿lYCcG d$wyMk6%~t(((ݻ>6wܙ3g>Ž{2e DU!}2/R1B8QxtC,U7hSD0 <.J]*L~ >l 4FeXeb9Y ;ˡB^խJ5eM_|z\v)^dEcJJjH%VZTug޽2UdN%U$(**Y,(#ʿoQ3Jg,-ޚ,o'U\EAWzʒ*oBfj6ӧz=>dfBb#ߙ+Cs`w~!+#1gVho%;v?MXqbΕ;w̽5fؾ)9B RA,n$1{,zBCЗy)iǬ%]yY0&}J1Y.! CiOGC~4U6HdQN9W D5´(HNe77*2 ~ը2KqB89Q7zgWoVh{~wM*>OOVUJeYL!F)]^^nZԓOn=YPcF]>SD-+s?z}2s= /- -RBmZvWaĂ^֔^wAGyniU$#.MXV0Jm'K^%=mk<+H׷ܑ[+0>gJdy~MZ< 2D(I0{:9Ȓ˼vk/wT24 J?屍p-蕥*2ӝ/O<>뮃Z!"qnoocd1Q W7$($dr|F "걥_8 /_G}D@1d28:UNv.ϓ K՝#_lD#3!@b#t6_~ _u} 5ό"iK$E.EL$$$(  ((;whFg )5 mUI8Iݻyh{XkLѽYLSmYR,,Ðpir.BflVuIBID,ߧٖy- %Lyoܸ41!"2H:e'L,^?ڭ,=lo@+++oYWܻ&ʽ=>J)G}S눈'NdGϟ?9 _-<(ea6BYhz;cO^pJ)3I" nEKs#F'E6"i땓;举Vɽ#0Al!ns뭸 > "%, ◾0 }2ggٖA08q I:HQFQP JL쬘FEUgЉ:6 *kaDHD߶i%e^4KEҦWy4(&2M?-կ~ƍ a2S(,aLMgE$пMkw2|ap$=y^e?^w{1??'><#st) RJ'Y~#(Vs)8@U;,O5/am %o{ů!US3d cqN&0 (%aM%;6ʂ$͸mTj.qm@to9Uɯ0|S 8Z.2l U||YfVc֗+0ʯ,>2;Xr(ec ("QdRI{-/pN89~zG{%Zu}G?я~z~ oxOO_|衇$N*;k"L'c'% O ߈hEk zi>q *DdYБd,)Z])-Z>1tzw48*r܌67ߌ?OQ4:ݷ;*y{aDlIۇIDhUQt 0K>`YE$"\Dx`!ݺmWСۋhTԗRf!W+nf7> D<#`ʉcz<БVXov&& !K+,h]'qwG`J"ҮeXL8{.FIHT M%AZ^jC=|ѳg,Їӟ>syp%c% y?"<",v8}{μ\>ڻ#QuKc Gpܲ?Ã7Z9W\)mU2(e*pbk{$rN D8RETÈĵAQ_bGu%oGQ~?g(CF-MN-Bu߸q#orI1"e>EvA$Aa]Jn vWJR$("/_e~8}"ƒ)Ssd2˯6SZ^+n\ 曓F[Z@<Tť?M347A'w֤Z>Yc'l,<J)aeR#`^ 3pqD)Po";ǖJJ#G JӞ]48'"U#2 !"ql" ;; OLtzBnShv `۠i8e. ʂ&D>&˹0YzuZ=yGN BZ‚qlec?""dן BaٚAB!$AF_ oǍmQɯ "& "gUD58XXD8ݢV !T_b|X,pMBs3հ^6,QiNJ4boܸM|-)?'FFymL"B4N}7aL?Dxv>NBrF"BsjtzG7/pNgb4꺥y+Z v@R8/W*wsOV9k'qRDR:6t/:6="#,U޲(4ϋjjdkK)~򓟴<3gT qT>S.jajfFK$J|N2(PN6"TjvPJ)혍 T(ɵ BUUIM؝^ETlm0$) 7I>>\.7۩Y_ʯ,|/p8+EҀRC=tF Л#y'p`ú*障 (ciy>"L3l1J#ƘI!%Fe]o^ {ڟYl u9t|Tsqr"ء oo} jœhB@@^+Je~1wj{C'j`a.GYIk YZvTdI-FL[qv72FYTLL=SGCE908QbRw!I zǷѶ(2DH_%ۏcD^_0YT^̩Ůռpnm鲜ab#pmȖM>+Tܣ, fJ$Sۍ=7C_5 "xĉ0#IMe1ሰc6Z)կ0TG{g5>"DDn$tGo#7 ߒGhqE^BA "B7prx=;9؊8"/-Wwc[ͺ&i`QUA@HG\V& KlgE}ɘRվ`r|+cieeA"Ed!,K̶dL5 4矻3Weh? M6(e0fYH*s<Ї -f՛a#$p]/}/' GVXo=/RD82rR~W`En蹾"4y$i*BxdfEl"r(0(ogbQT˅U2dtJ[3H58 4߉: r+ZjqĠpw #mn6 K|SqQt㻎ij4 R \w$LYFeg9ZRB\Ҋ` 2 2Kq= h$ϲq8F w;qjV+|C-ؾ^xM-kf<CP59΋mY"X2H_#  oc_PQL1AnT|kD<'qM]?b=K6iD""K.rXZg`)첣On~?؞vUiYe GYCbiY(!0uu, eݰXK~0x~ 3v:nmQץC%čC;~aQN ͢ ,Ҵ^L77kTL"t,24ZrrB8o;/в}\n_OҒ "d߇%dh^؞95@@g~?_Q:ZGO?MS/L |QJiN-G78/ڦ>h8DjPCMo&ah "n8'4YnjoٸǯotD8 |"Cj#Ys6~DXO@QJ/\L)?wjRj(H eD|o5j9>yR@$IFl0(ahiqK v~Smbw&noSpmuQDyJ[om<'t]YZ-P(gPJzQ_9Rj"mbM ׷ K.ݸ6biFH#Dn~xs_ʯ}ckN6}/zzn`K8Bf󽱍v|-3QrZc6R z2ۃ >9z3闟?G,w}`ow&xcRnvv|lz=V ?#Z\gmYh#"Z$feF-y?7Ak$klB_EBEA]OчVbZM]1/,_dIP,opg7>7aAS a/Mˎl dkÏjy#5~2'S r!!IYDgB](d尌*'58[s9MQJLaƀ:=RJ2V,y Tr5xli#7/|fCd=j(ZQPX(wG>;t "LTJ,"7dVA?Ep}tww׭])׿݉67O!8Ƃ^\("=G@D`#" bY 8V63JM)5K9:7M2xJ) bQQ4'/p?9]I 4Uҭ"Wj|(E8GB:58( $q<^"4yQॡMms{bz] EAPչw.XU;fR bs¡vysȽn~J{W jѽ=u;PhGHf>e5NXmz j_P}eȬi04ɢ{8`O+n/77i>,o'5 uJ)} C9]Dr4>ЌeW٧ϴRJ^LҾZXk vxID˂K(eƏ?]-ς yYPD(IP+T9!L5z?od( Q6dЋ[ݹ_<094@1Cxl# BtL5~) BMۇp_EfME!K7yG~Hs]-llTNoO~&IP*h׊6p,>mFiQ,EQDUOq"V*m)z0iQFoٸOR4gھN "K<4u! -l)"9V7òhQ\:^={BKm0ki@ {!tWBj#DD]9('v&3"& YaKXBEF5g>6ܡu9Ǣ$=߉%7d_Go޸ry?t{o=׽ ؀b<{!Q w/QERIN|)2hGqw;bOK.T qXylB|W,%F3NyPsǗ>um( "6\|׃|; No"bO;Ε[OZ6܂!lo"%5[#b.n2rTIq]Gyg(GdΤA$y[ux~_p * T ugTn84 motvo?Fw.|{&CH#9"̮%9~/xZ!"iE[oN H8bu ")ZX)"@˂HPQDMôUe= M]"i! Lkv'ʛjt/&lI!J2;fl,s쯽CZee#/}#Z :$߿QC_61`T>TH5U0D;ʎFQ1uM6ʹvҼ |R <[ith,ۣ"xtqf#e0Ԃ(plNxK(b3H Ւ0@5FW?.C`.)q([ 8+g}8p,2_]R; h= r[ lQo4hg 3N_`rTIya9 Sf"2gR],",(~@ؽ*K\R쪲,2o' %+9jyTQEMo`(6r/~ؽ?pX,Bq(3COx~ziӸבkI9T$*lJ%X[ÝmwrjU_;tz{9;[m"PJK%LQ%pדnywr"bZoO! R'5b8P۽=HλJO<DgkEBxYxl{=Pn15⩍q;gQDZX;}Ǻ&3D*9/5ID(O f#vp~EA$Bx3!^3l0gty{a=BI5Լmcp8?Nx;ID;c){ր@1gp/]zsw܂S0n[ҍνCBe8AUܲ  `ҧ_b9eԪn}`0sv1(^:=@,'ɸ"a0p BD4Ԃ)aD扇 <)ǁ'G5^2G<4GP _>TK5R 1^9ݷ7W0""Xx}$Y(3{AA0i#S?,M7JVVq=̖ W,?rbCY*;{u03 @=WJf:e-׷NdY՚-HN)t\Μ/~q,bg+!@R]y^3jؾ[DZN-{P.L jqj"/6z䰅0PJK%pEi O>(eB&EN4rEjwH0*ZtuW' g*ZXN9M`cN ",/'PC{:[YBe (UY}Cp+NOTozmL BB0Cm0 %โF3eJjRʊ5Vr+$ NQj\XS%0|^D-o $Ȳ1A7/|{w#밲t̙6B8t[ޓloFT5jkNZ ; IL0 a12҃# P@z#VA}Ib-mcz TZyVRU]Q ,@,sBAp]P:o>W_su&vAg(KJ=~ ήW} "8 !ٽ sEDߧ&0AP,"?͡Tv,n GfyrB9!в!B/pOdP ""&TʂCl!<j^osgq~3+eq"6B0 4^׃R e Lkﭷ߇bqqs؅oYКD% x^D86œZ܏G8LwbrF !&;h55( %d[ª %c5;tTzW˹eYPzow\ۣ7`~"/Kl{s+ufzbuaϞf+tΰΤcDX,"qsJz!yNPD<{^Q˜QHB 1|>qw-;1ͯV̳Tpw~ZV S%5̶l6)a3vikr"Ta t0)4@}Ge$Lg&sal*+_ILw}~ \3nӥ%؀xz,8C^9UW~Mozӽ;}>vmwyLG gmq JonN-C\rd߲~m`ceHD<P`;_V$b{V|uiZ QN i\&7K4zuW'꽭88W[oZ33u"ht:W~hYP*vdU%ԊĚr ^4;i#A@i$00-¡p@UE1E g!\9QvF˅pЧ\-@Ģ^@g0dxIDA#8 Jcil]Y58i CZFb+!ps"B "CpMuE0HX"OX s尴2Zmn)hne,kyMo>}/(ٖ&u`%\<ÀZUnCTU1aDl/;j=G6 `4 $ gܤz?6?x?#`%'߽ @KKq_u}-oʡbD78S?G?wFLZ@ 'k#D'pw\˅jaلr9;:[[4&X.U9#mۨ54`?"m焮dGE)`,r~}"Fkt zowmCjPo%n,kn9R]H;fcZ 8O;<e鶗N3E!%~9hÈDqH" ,DKYFIRAB&ҤT⋶- h YmI{o7[/d[6)\ZWpJVɧa-mł0oFEDET@DUE^ fv̺[[ԶJk5n;~Dj;QEgHt.CV"N"C_ňg~gN6=j__4ǔeE%[9QC.u^$Fyj5dGJjō( f}/pcdRA\Y}o)-pXL= BscNE v|K{"k=|a+ܸ! 0_)!BXZv{44"^/q9ތ7V4AAׁێͩppTHdF6,b~8e.% ˅a;Y-^z2?͝q>*_{"\ QvA-ϢdOuYT۴PzW,2$(!7BɓҠ2D x^,~Vzbo\TTe7 Zn='AF|O+W>0NW`yIDeqlCzefʾ.|JPDC"_!"t)6;6nw/4nay)<_Yl[dvrN>ON8qUsy00"P"(Ձr dI^ q}=ANmi(^iώ&wtEE~|GcE^++4WշͿI&U !AXR,bj C!q^o}ڝY!sYm?4qN7Mc`eۣ?<7+4ڷ{ CIљ0 $7=3~!lTrY.!<ZJq,W{f+dvV ze`w X[jni+_=woZ&x;DV֏;fg}qgj5 '#輳8rN}9=DҍΓct߯a$ g|mr@pcC~?B'W,5Z'ʕ\6x 2Bo&s0pla"4B|Gݞ>oKze?mcET'}NK<~zNwLoO6T;ށ<$5Mur J_ňpzEX"b x@@$ D K)elԊٖV9xC\PK88"d|4( ]P Zs*B'e#ҵ 8z@[ӣ2F !Ge=qre"BD8}/^ܿ'@H=W+8ب*imzltd}wZIz{jax(F++FJn(&X0 ޹;ۋYd^waQDYΦsxxqh47-Ù3\qxsnL5:Ge] k ,/c=]2$AoK(ju?!dhv}2n UEn5 8#"9TٙJ{O<Ƿ ze*-`c#Ar<~+ +r}7@snr/an4B;;;?~{o-4 P"B7]` ~h#:Z@mvɝ3[ 3<nXs~yqaĕE-{4Ev" 9"F^?2Z:1i˅(KO^kaVE>J3=y776@?Fp4\a "Dߓ,o{R9nR~ BLM3YYU6 %p~rЕ<+\]~GOw,hm8ۣ]SšO<Ƿim ,qע~u|xN '= 4c "xPΝ~s'csaC,GԛyV'+FrvKw~۟_Gֳ;ckAlPRb˶B!'MD,].)KrjNշ۲1ڂ%8v i%W :JisR:V>q~i8x6|InM4*Fm[E),qx׻z zx=e- U۝LiF7tm Ck ]ZίNM_}E/n yr糆WZWWF!/כ^}_}=kRIIx_Fk&WvG5E&[[qϾV% N\mcS/\}ߎ n/+r/< OO|{~ԩSƵ0ը<_sqrg[&q6JVHE $dd@#;sc֙ g_F,"ygvvrf{G3#$=g~}5izpG>JuZ[,LBckBdE!Qv d7Q% t: s >#M)I("q<Aa<)8@ ID1RJux]+Pq{Dp ǃUi0pє\٠0>b`2 ZY!$C a~(+±@r0"sīODO镩,"*b0'x>)ҥ%TBi}>bA+yTdk8*GvC2 ǣ>p J+B&*@ssji_+@7L?4}Uo w?p(&tD#$@D"d!̃snu˭a`D#N' 2QqJ Ev].< (U@diɢnX> n4y9HL[Oq`1I]nL6D#1ȶg6H090op@%04!ꂲ2d2!b hu25'Sں< qSЅ/w7;_|Q!ۿ#XБ#t!eі#nP$au+FQ!Tiou[~^Odk4=B2I& >!adׇʡڱDEhz_iQ #D~@ɱ? 8a̫ȉb}ŦGwNٌ3b4JTv^sdrAi)x< [1T(:ЯF#Tƥ$(` ٟ8HDcQ#~0 `CKq #jѱ^IA8LV`24W-La $D,"AB(S'x2&0ۋf(+C{AGMQ=%$غȿLVL^$Z !P(`BOޒdS@*%GlRH{ExhF *+ezdx1@@l9  Od6#HUNԄ P _Ēf3jlSwR;ۋy!"2 Dʴo`(y]"nAbq`^e0`ti >\V0&Iz7VWz%:<"]y'2q̘A= Owa!˲A0L !$˕Je*X$J3Fnh#z=rqͮ_iksɖGbDTΊӁ}(6.bRF^Ǭ)I I!`Y‰Le2PȩdTt1eBN6wX,DZEc)*3.S;bG1HSVHVBR::QN0U."H A*WJ%Ǥ Rq(ˡ6UHZhvZdȾk"X">OFW{DL8,rJXbЇ"doPBe`xP@r@00VrMxF|4ɧ1S>H? }W k*1^BVU%[Zhc0ZѮ.P*sۘBfs)4'RiBd!q*ee5(Bh+ ߾.J%b%Zw8^0ǂ,Z,I$B,Q(U I?pv~R!1zo5Ǣ)\G'8u8W] &l:QL?sﰍcl 'X|\Ƙ~F8BkȢDDb{zlR6(_$NऌlVaIDZDt0aC1 ͦQ5qXP2&˙Q@dLd CKF< CâXbz*}l8 /7Xx'ݘ0D*2 v|w;r>R"D\^X8Hd /BPXw@H玿rˑ}aNvgxʈG4őxhlp}"]9JhBbNNK9 sP\ &B@L`* l <Ė1E~ZXqhD_nՠ6qU,&"r8 YzYZn>~dѠlhḁ 8@:CSocԄ VP"Ak{#G:DyףdT 8ҏ bH wB.N$pz b#TH8sGyC:ΘZ,`2A$osz*R˟"{-R)ZE)T \B߻̩3Od uH'eo)FD< l&@K$A8 Id4A4;3PBHʴC[bTP`{:G1Df= *7v85˜\Z$4vO#AfYQ 8±@yT 2=HC6J(dfK'LYoԿ"̈P*! %af!S6ɨ.AY~\qRydi%!U8p(U„rJJRy<4s25Ԗ`GX[HD!A9ͰlYjpt4$sp$.$d#n@qYupujp<ͽOhl 2p,eCR{$W-<}ru4${mGu\F'D25Bv 9"eL :EyRH#~ҘHW.-@DRij% -Q+tĤY8f3gR)~( ^,R !+K IrVɵ x$iBCѪS0nΐ KeF͎j;8(5b0C $HiqJ5Bɘp% ̛ wLVR r]HG]Ч۽Dn#AQH9 >FH8 ` f~DR= 2BYF%`YɆu.x͆p}n>_֦aeZޮ q_nVch0Ä! 냢"AʠF!A1vlIsAz+󼬇V!ɉP8?d~ ^5 ۚ\,_-h:1ǁQS7-NExYC 0dXjYi ^8C )`[$Z*S#\w &Sq4X}!w<%+B)9l`ͅL؟JE!)KloYSLQ бþ}$(?% c(h'߮6Q3 d ׇe2&=Lr9c[p+F1v*0P=xC, P*mP65H41Fٳ$&u1jiɬctZ%TZ ,FR%('+^/H$Ĺ4DZ6rav>FH~ II"P(P!8>GDfvI?bm""jGL8]s鍯\:pMMDmT5P62 {,F|& OFSBB(Ƅ5!d0ƙD*%t ch#~ +0Z$S'Tj@F '^J!R"M31T 2|u:@9 h5Lk'~19E"XNi"!cd2N0 L̝m@[4 E qxB#!4cu@j!#[,(6BZcLHa@At֌xС!ۍZPUp׮c~Jb8;X WZ|!;/|"g z,xR *:y\ID>-1Ʊxe$idjlfYR'r"@.Q濟caJ mgswJhq({x beR (7˅=Z-+H6ncˆRZJ<>!`&aAZpP#ġd@QHMݎ zꩧfΗ W# 0 '>[;z+S~q_p|*_B5?c-<\B"ƅ)˥ [?LgS 0,IZs#*($cߟ)MiK4 Áf*acڅ?4 Q>G#(y|tUi3rzP@43XR @aW~Sdjb3eeamVw_ۢ-/V j#$|{ 'OL+ss|J !22O&Y[1A.tCfy)KX} :昍WRQSDuDca^5բDTmՕsEn7` `kDq6ӨgĀt4bVF c|}ۑ]WNe !B`_S&WnթMu V9HbT$s2UH+jNBؽy] X!GKYmKþX1N'X7S+r,pX"Pt:&f?&a8~A!hjg q!D:=Nd# RVKFnxPBRc8v /\HE  ٫}qݍ#~^N(/GvCeO=]d-ؽ0Sϼ:zqBp#yk841Dv}-gpƁ@G"87nV +$e Hl#HՅɖ()7ʵm"$0mVoތˁ$(pԃcƜ$^ϰ*egd"(+*cfC==H"P( ˂LJ}c aLDqB -JKY_L`єm`CQpF D!UԯLEF{{qq1`*\4% *7w2JD}`R)R*-~ …czEML!IYH,fIb3e4Y C!lFr2'8ѷml$0]R)z1cS~/ţ &;YZ}\[M@̄@:;qi)Bѫz=$υ3TC; " шJ\׶e-h9ֻgu+F^$8&D Bm-ÇSG2V!FVW#/nD/ŜLmڗؽ+hzRjTeb?@Ujccxm]fDldTctu%`I? 2[Zk,7Hm8MIzm먒 `HkGb&%˰|F,㮮h 8GDlKd2 5cF.A;7sP.`VbH<*|6BR# cp@(-gh3 jK3G6AȦ?+3yMQñ0'J7ǎaz7`{ޚR,XB--f v;TZ??>O'[[ .OMB9xԌF$cRpKnpt: ٚĀBzTOM܎ ?h#laFIQۇ l!DrRj9c"02 #-.F$!TdKP\Bà!wK9T9le2&B2)'e#D XXjz)+2O|@FlRmj500E7lIQ`2Ar9%h,w<-g"x,P(c?!d0hHu{j#ap6لG_p$ϯ&Ӄˇ$d.¨&E[&JF]n3Uz_d>qrFKD}CP ebjna`X ` evk55V\UL&(-{s'V^pdjNN$k0\ЀkOٌ@Uiq5 (] G+.DǙz)H\? #^a0v,2=㈥Jx$;Rӆa((+C1D@ ]VH`} eGŌ%[VH`CQEc,e x׫Z#hҘE>w[7 lēx2iNr>u M={eErb&0b&R4CS,0̉i"ϒc| TT I|=6B܃z1Ơ4BȗUh,ҏwKIX12I,jk_yŚ`0 UhKD kj8ISܛr;qrFyޠr|> PSlm% goM jn>N҂kjEĉh.\]q- v8FJBF?1bcQw7xCmv} j 2LZуRkL0,ahvZ<QP<'Sm2r*ҕz: EBh<5jD z^`( p . MRVWluRdXFr?_:_&]Me4^e޿W"Bk Rj8O: b1()9؜2#Dei9z54(tvQW#hdjnCTX aj(+C۝nZ[1)I2;;S4'Gq8OAeai8S Y2 dt s2nŲuwc'!*WD+$+XX V0,JzK8x^^~@)+ i4P@_`YF*a~{+Z# &"hO6*mޝ͟a#  $^O=qShP#%"B-D CE$'TQ,J`I$]])vϒ~aʔԝKa$L%Í7R BJe{ :ƘB+m<;7O3lcuJSgKь qh w,{ܱe-5arbHQ ~MqC5 rgZ^JDקDZDdU 9cJK-~u˦xǍCqĝ;V㛑O↮f4qZatӁ74K U AnO$tx!BȢDCryaJJa2 --9HYnU%Y!>_p"lv'T*8-hD?ů@n7\LŲ"hBwƍ(NorRkr5BH0BhD# >x*Պ| D|- fooKD\N:JgBIҤ)JO]/&D$Uh B$F4aFbCw[%azK)BǁҨ4W9 騦3+ Ɇ~$!9m {JL1c[Z0 jL8KK.B¤L Lej r2#*9mt:H&!VL2رq)& Ғ~lm&SjeawofOpTEv8xLD{E;tJLUjoJ$ȣb*5PT3_;I9Ch m*ҕxv~+6t}fmNҠakcQR &hLRVNQ4W#h@.;S!vA \24Rۥk,r c޿?2ZdJ7Lhƽ5NduNSioo)m2G^o*(Y>W QՖ1-}O~k70 RMMvH>f zܙ`BرԷ|K *<ލ@駅ѬY1aMIjo}}l6@uHMyt))$BR:!˪cUyPF xT !Dr<BC[qwf w ?{ Z0>S^\ Ɏ[E%%pfTBR*b:#8|o+VP 2j:!`NJ/+=m֭djW\ ~rRL)RAooj_}[*/GC3t=_ȭ:hzR #XǕO%FW{=>?H$H0\)c?g<9X9E0YNPSir֩-8F򇒆)$,Ƭ-͐@s3 Qqf;pz TWe5e $?A a\_uu0469)zW#%LS4c٦:]w9xBls|4]-,' .|Qyo&DL\Ne(b˃ٳQe%zcp:!oA`}޽ˎ땔Qt5g7Z 3^-„ .]h[ټY3eg4 |\^/|oT`rBr.RVFS41TCG҂G:iB!LsPG"ۋR0W^n"`YoަI7iD2v.m)EMp onm>\DXm^)lLr]__{5WwIZ  P_wkZfCMM B@BBƲ}&W^RK ph:A\ᜟ&j5zZ"/g_[.s 2,.̅ a現?^n%\ʧrѓN\@} }7" 9M F4 Fm݊w:;qI #,C07cS"$W:)]([n ;f vp=|$-}_ssՅ6En'RS m91))ZAg{zt([~ B4i-~:ߏs {}åD*鄁r'qЛo SSG$KPf@ˌd]tM$RG.QM%fPBHRlIƍC}}rAk+TT Yub W6O T*G?ylFIJ0 N 144 /YBeE;1Q{LXn,qB־< ͻ81Vdg[KSZ]pn,#[' ~szjg瞣ozec!W)!R/RVN!I#2t焖!B`6`/\2V섷Ӧ{索 ?cQE1_qغB]NiȦ@5ٜ޸:[o rH4¼A _*EK( 2cέ--OqIq0ƵH*Eyr(.L`dRj>jjB?IA0mZO xj5KԴG}մieFCCV+h4؈3CKB;flLf\zj]EBÞ˲2DP=^-,[F]ye;#/I=1Kg9I8Nw5f1N|J)eaBE uǩєACp]5P[ᅬ-;etydexR̜yb#D#ܲE8\dP|BMcx]A"7 &I(aw7߅?PSӦQ8I/K ܜJkI}!)cFGM衇JryAGOF,h{sh9Q~=%Kۓx<ǩ UN(]]xF.hno`en{%#B.+؉B`̘AgorU 'g lꛆR ho[oEz !rrPѣmތag<?q"=[t9sЌv_|!ʏ3p V !PVZtY:WA(/GwA]}55zMI>n :hY^!5*"N|64ݻqg簞2k\k=8Zg\f3('\$!*7R2ʕԒ%d$sWG>JكyO$ ^NEƝ|7O&G{?}T*˘G_TxsAOMpBA$mӮ&4grDQqPWr KKJBp>zB! 9a6M ޚX2 F^K=t}71 a|ce᪫N(T?Ɓ9ImTef*?Q?(#8. *O4 Q8|GP};O<~G3̩жЯ~E?? ӧS'̧=s@XA}bTQMܔ(-HؓZhէS@ !:IBơݻq$"qNL pT:ľs^ihw݂ ~9a6C}qQbP^~^( jtčߒ2]UeFZ~5TV:V+`t9gfHcc9jX\vulEɓǩZ ѣ.D$[Ʉ_Gj MMMv/2׃?5eԼ@1q"zaj3k N|UVm޼y&iŊ<ϟh :%`6^r9%@M0rOf0)E*g~hov LH:3GCW^-/A4oh/˷݋L<(=}zjW;X 551D"EEp\y%N$ ?œdw*ʯ{XxzG`uN%xO$cfҒ6UOxDW1˞pH&Lg ԙQQF555z3gwmX^|'475A['Nmw[ZN&Hr ôiӦMv,,,d?݇|H&v]Ӊ`۶m6 At֮][ZZZ__ܼjժ{W7H%wǍFhv\PHPh4t6qgqjF`H~mmmVU܁=z`08m۶;vL(cXwwwoooz;)Hhv\,.].׶mU*w>r4>6mڄҧ {<'Hofs ?zEN=cÇ+a4S{D6mnX ۵k7{8Hk4UՊ3m޼l6_fssFpꌭTh O9%KTWW,K<_0ޘ1cnE3G*//Z+Wӟdt:O?}Z,+***+++**9s\{˖-;wn0$򎎎iӦ,_?f۲e XO0eʔ{BѼD[WWlٲolƍ4~gD;3z{g̘1({_y .['O|x<1nmm5jԭ:s{G<wF+++l61ްa%\rm]z饓'Ov: B49sҥK.]:{h41~',YrwyK.bG}bŊΙ3g?l6ێ;0ƉDK/]d?|q| &b˗WWWK$gtࢋ.2 ntL_nw0,))BC AX|c=F䍍s|7۷dq!x<&,GD ƍD`$YjUB 7pv!*+++;z(in>tPBHc֬Y+~<<`^x9F1;~+z.o9sL^lٓO>IyWcxdEEEBO|`0fqIT*–-[͛R{͛  ڵk̙9ǰ8>;;;G}:/|DlڴiܹEQ5w/w^Zm27o<}tq?3{lr͛Gu]~j,w}o?LhFW}ݻwtk׮%ه0 ?ꫯ>&wygɒ%98 ```0F8 644ѣ>!e͚5}GGի \.BP(ȿz`^qb5A|s@ `4rPL&2 w H]w^lc1`͚5Lovhoo+} i?3Zwtҳeeeh---moogpt??|7c90d@ X__1O̙\!af͚wr}"cՑ)bY~ԨQGyw]xx 0jzѢE{:#G;S5nW_} 7nF ]f6~i4ݲe֭[\s͊+Nƍ= ׯ_p!9{ƍNß|I4---EYp EV"7ndΝm۶m޼9}g 6dC.\`Z?۷K7,)H6˲W^y?_/6"}'x;ܹsƍB+V+x MZ[[W_}5yqu׽۶m۶m[o|rxWw/߿kiiQ*ݻaÆ۷袋`ҥ|gϞ^zoڴi`Qr# ۶mۺuG}h>Cb<=8m6mڶm… ?#z{ 455yǓkijjZfMMMn'Lׯ_jժf ì]]f L5kyfm͚5~͚5ƍ;Mx:oz߾}wjA@dYYYJh4&%%l+W{* @iiiNN,Y^^766pj5ӧO|>_G<uJ0gǏK,Lp\6h,X]yJUUUpÇ=HxҺuܹ%I]]<m;BVrrr]]]DDtMMMw0  X,Rn2 hܸqcCCÙSSSW^-Hlɇ'͛7J%r\& fgpjuBB|*+ Z-㽽yyyA {zv'IRѨT2 zzz^|aÆ"l"ȴ /^,(JNNvlE1}y"ƍ 4/,,lG~:ɜ"˭Vkvv661mxxX*I<<K /Sb v;̴SRRRrss1 KJJMEr6 3}|Џ8 ‹/|8͛/]{??aNh"@PZZ8F!jd2Y|||FFInwccH$hL&jRvGGG].Jw~;;;VdUWWL&QYY ˻nxQommAAAjjz{{K$Bt:t`0?(#ȯCih]v PVZcx֬_$L$Y,T*rѣGp_|yyyyjaaa``3g> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 372 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 371 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 57370>> stream xw|TUMK'@Xa]AW,UUYVqU)* NM ғ7 äMIf&?grʹo޼=K!@  i@$MC !@ 4MÒd@ e(b Ж!!@ 4#$Bx@hӐQ@ ih(H"$H"ۦL@ *dh@ m6 0ӧOxOMMJĎ;B"B+ŃB,rɾ} DV0SO5jvZ4iV>AaB믁-{.,???%%%'''++Ch׮E]a%%%dgg'$$ݻcǎ]tsEQ=Z\\թS'Ç 222t颪jII$I{Kv{nn'233NGnl6i{q:YYY3\4܈v'$$3gl6[nni}u}.W^qqqx}3;޽{].,EQÇ^xطuرc:t 񎕕{ .oxq_e/pB!C27>?0%%%++{ڷo?iҤ̜9sȑv_~aÆ;55uʔ)oڵѣG-[@tݻw'&&nذG2eag޶m=o~홙?k6| 6̚5뮻믟x≑#GVTT 8:u(]t :}8mf͚k׮ݵk$I۶m۰aCJJ}t:zݽ{ovm޼y]t駟|/㙙˖- 'Nis\bE||Сݻc͇ '|g7o~3g\wu]v^w7n\^n^zftݻ/_ɓ_a(8n-wuc1ѣG9{'x4%K\q)2q[no͚5izΝ;YCYVVt:k;aNsϞ=={uݻ( 6o_^^nfiiiO8a۷o^{Ϳ}iiiFFt'N? ƍßo?4o'čK.>}:{͘14޽{oqРA뀯LKKK-֭6ln7or\YYYeee__g~ǿqSSS^+Cm@,!X`پ;V'v  U /oر~Q%⥗^o߾aÆ%''w !w뭷^z饃4iѣi/7o^\\˲eee,رcĈ<ӣGxӧv{NBvvԩS?QFM>=>>[nK 2iҤk6Sǟ322qٲeˮ]ƌ >xX2//߿?>>7xA:qQFmٲ%77n]O:(JNN?(.kn{РAW\qŌ3.ݻw?X+2''S{V1bĢE|UW͘1k׮$˪!DƢ6 )) ΂Yf[y^Qrl61n(jO>/2{lF?άY֮]]RRҧOf 4AdYƟUUej^tEؽ{?;>[pSO=/̚5kҥ^{m?{zjZM6sLE9w}7997b?eJKKׯ_?s{j9^_XA~s^y啦iv!~cǎ}7\sݻV_UՊ?$SSSlrO?tĈǎH"#+cx!BhǎVھ}ӫW޹sg#$a8ˑ#GxԨQoۡC.zLȑ#Ǐx,Yd̙(yÇg&++kŊO?tyyyJJ7`q>x⬬?3g0ڵkk?W_%g={ܹs?Ν;7++ u[u놝}&''Gss=np˗)]tђ%K7MsѢE'Odꫯ;v,0+W| EQ:u!C,XO̟?ܸq'O$ɓO>dee  ,СwQZZzW,O\y0x`-P5tPKJJ~+I9C$Xj>>/p6믟8qiJJJ.;wo>J!P@4"馛馛p&))iYYY0xo8C Bme˖=3o t]X,GIIIeee6m׿'G NUis=}" #VK<뭷N>8r{t]=eYѣG;6Z Y99yʔ)񊢬^zȐ!}n6m^}_0SPPsNRCkq&jJJ<i2dH׮]I*Y B 0Ç8zh<ZrmvmoībDQA ԁGZ$B iʕ+?sǓ=cƌ>}|6l禛nۗ.[lӦM4M8svj<_t:8n&Mׯڵk'M4iҤAաό3.2xWn݊⋠GHhM[lҥ 9sΝkZ 6lؐ9v澌%Vcǎ׺rׯ1bBox$<%..nĈ-5nLQTǎ;vX/G-ڵOܥSNMәշ!h2"4ò$I|#MOw8ݻw?y$'O}5'ZBnw3cX5rݦinݺ~߿?p w>q#&@ X0\rG}Eq޼yz~{۷/EQ˗/L"5#GtZ 4gΜx<6YRobXMiՇ5Mc٪i!k֬߿xb&3,\0ҊD#1ȍEQ5fQ@ m2pD 6 1V D!$=3ւNbxh@TU4_8: !@ 44-Z%&4!%Ν;MӌMleTMPbAI"];"lw': |Ѡy]lo>J 9ƥ_yPB .h[+m 5WrL7Y Hn:CbDN9k,<}5!2ʥ^$͜934BM$YfhЄȖx5YTuŧxBߞa]FQ(1ѠI$>A 4.{w Db 6TM 0a7>lr޼yaD&DFTuA1BY-Z GD& d!SPa&2M6L=ĽOh?d^24J .@ 9L'!lόfʔ)dyzB CĉU+B CÇѪUm<BBBS4MѡBجZZ S+DQ@" ,pEQ>2 C^[!J !TVV5 p6GhlV- 1U]6ny0"!QuY y*7o^jEhaOhVMZR* yM"(Ei> &sa({(mu}۶mv[xo,D*MPߏWhkk%7QjU/Lq}"(i ="Fy,D!gFqe7㖝j Vys,EG!ߙGbA%q j4MoٲE4m4E* ,"QO$^Qu( T " .Əd(Ҫ MbcxL(Q#k]Ru%JL4h IC׼(jwqǺunݺu۷.DG-&G(Fۦ?>0ag=ˆڤn'S<2 3M"'gP!- /^;hD& l5M9rpܹrbŊԶ4E* yJP4 yD^&~j ^t5o;R34-`38*Ϡtg aHL4hެ_zc㴭 IIQ,QUy8B}~];"9T]Fc;ֻfUG7T_yO(N9`TBMW$k4,D4W^}]wMKiY5h۶mÃM9n[GbΊkى Q֐5sE)6ꍺsΈZ 1°(-MQԄ &O_ׯ_~4x4S]420Fh't閝6!pz~Ohq<=+}2(0Pv~֨[5]эKk/Yz3~eY^bE{MtaXd@~~io㸷~[ YFiqUEo*=+80 Sx M!(`ҥ ,lIII Զѣ999 !k׮W_}!0L>}vޝxʍ t" #f:hC-ٹM!77SNM<p2|aqt[ɇf{\Ggj2C{v)TѺw4M3//c"ai(\)۵km۶^zu9///(q .<|!C$I(bU@t:l};:2 HLZGaG?hFp/nbp͇K.LKQ[G2iF`$ !FlP!%8}u(ҥK.\taiY-:}+/_ aUhyu[)&N24.pڴi4݀RWaQE~K!di 5S,rI[2P/YG_nTՔ V᝾D[Jm; 0yF+P|gۑCfhߨLxV uNz,wѢEGӴn޽]vopu麾o߾ ._ر#=޿޽WXQ[|*xh4#B!SF@$Ujj-4}|R5tСC}dY4 4 WIf0[bE3TЃBqeMGqɚD&+4CL PIxwIlFH1fh@xA?`Su0:x*oP# "'|Q&, -hX,@t#ڻw; #+ CR-/epQ``Y beMYAyy!o(.MW?nxV)+flhzȋ7n8, ,Xǎx;hYQr"O7{ɯ<BB(*>>}2a?4iK5/PTtOגj2n6U{0 @ywݏ6.ЂqÄB]WϹhή,Xǎ5&]'q͛76&`EbNTfBAjz@LTk*sD!~tCQe~`K@0<~6fn_e{fT5Df۽3Y-M= ` 1enIIkW8~GT #$CM-B>$ED}YRGl従[<ӊ !oA K.6ouG}LdaAS@N&^I%?Eb"5z<+24SwR2#lI2UR|P+6 ћ,QU_~|_~y||],(REIJx. k>DALN6qTtBMW3'g !C37}_FFeb"Yu ߷?L薝<+3OY9w˵&z$X--@h/IX'N V(:ë[e6?{:Q*ԼG_~DUGv򜈇z*Fԡ*'Ÿ:fPxdUKyP@,WGN$-##o _G1Bt6^)ȳ6GkU]*Ч1b5\lٲ_7|_|Eb>+RCxxu5DGZ {Dh E9gjU|yE[*BQѣjV:=B@R:L ?~K6NgBZb*;- :ˍ Ey:"34fs _[whq˻dQ{$Y%#GB& d5#leBǏ#09/=0D!%Ko˖-kJHQTV UgMR5%#c!E.Xj0zA#I?USQoGqYn}ůPл7Ur:lV1?7deG9pdG%lmoR .((*pt!{eWmn{Wàz;O[mH몠 c404Kn76aMӬ8:04ɆQs֨X?ɗU/HRX+ut;vH,?{p6#LqSowwE+J( 2&HO&TR n7X`8*6fjK7M3/// ΔEBr!ը*MQ).tڄ!5Oqĥ4 V=U/p, a\zySPI ae"J]"*0T=JN;]V*dT<'bg1B WShz)Ӟq,U= ԅy]6c$UyYYpX=1KS|A>u.5bTE@ uBNoA7pm!Թ$] l6l!%Dp*a":[HJ M }#lD3h<\iU`O.Pp !T4o⨬zl)ayV"LԴ@0xBWgi7h2DJ]IT,g/sL=8N r VBz3 çxq_Wel$j6nW5{Oendi (g:w>g) P, úZU=j@Z=³2n7T(-z,ᇍET[O:tDZ)S;${ 5 tS:*ѷ)ڄ!,sJ`-hhT$^=4MEl)ayV;_#TuEӕcuG74PuC-,hV${{pRU$8:upE \!T)&DzCcO5{e>X,Uu➢# U=xӄ !6ؤ `U]> h^E8q!Ye- Y_CQ/0vzwOH)#Oac/<Ы\ eiĪ!4M֭;tPZVV?գ֢"҅e4]Y޿lYCcG d$wyMk6%~t(((ݻ>6wܙ3g>Ž{2e DU!}2/R1B8QxtC,U7hSD0 <.J]*L~ >l 4FeXeb9Y ;ˡB^խJ5eM_|z\v)^dEcJJjH%VZTug޽2UdN%U$(**Y,(#ʿoQ3Jg,-ޚ,o'U\EAWzʒ*oBfj6ӧz=>dfBb#ߙ+Cs`w~!+#1gVho%;v?MXqbΕ;w̽5fؾ)9B RA,n$1{,zBCЗy)iǬ%]yY0&}J1Y.! CiOGC~4U6HdQN9W D5´(HNe77*2 ~ը2KqB89Q7zgWoVh{~wM*>OOVUJeYL!F)]^^nZԓOn=YPcF]>SD-+s?z}2s= /- -RBmZvWaĂ^֔^wAGyniU$#.MXV0Jm'K^%=mk<+H׷ܑ[+0>gJdy~MZ< 2D(I0{:9Ȓ˼vk/wT24 J?屍p-蕥*2ӝ/O<>뮃Z!"qnoocd1Q W7$($dr|F "걥_8 /_G}D@1d28:UNv.ϓ K՝#_lD#3!@b#t6_~ _u} 5ό"iK$E.EL$$$(  ((;whFg )5 mUI8Iݻyh{XkLѽYLSmYR,,Ðpir.BflVuIBID,ߧٖy- %Lyoܸ41!"2H:e'L,^?ڭ,=lo@+++oYWܻ&ʽ=>J)G}S눈'NdGϟ?9 _-<(ea6BYhz;cO^pJ)3I" nEKs#F'E6"i땓;举Vɽ#0Al!ns뭸 > "%, ◾0 }2ggٖA08q I:HQFQP JL쬘FEUgЉ:6 *kaDHD߶i%e^4KEҦWy4(&2M?-կ~ƍ a2S(,aLMgE$пMkw2|ap$=y^e?^w{1??'><#st) RJ'Y~#(Vs)8@U;,O5/am %o{ů!US3d cqN&0 (%aM%;6ʂ$͸mTj.qm@to9Uɯ0|S 8Z.2l U||YfVc֗+0ʯ,>2;Xr(ec ("QdRI{-/pN89~zG{%Zu}G?я~z~ oxOO_|衇$N*;k"L'c'% O ߈hEk zi>q *DdYБd,)Z])-Z>1tzw48*r܌67ߌ?OQ4:ݷ;*y{aDlIۇIDhUQt 0K>`YE$"\Dx`!ݺmWСۋhTԗRf!W+nf7> D<#`ʉcz<БVXov&& !K+,h]'qwG`J"ҮeXL8{.FIHT M%AZ^jC=|ѳg,Їӟ>syp%c% y?"<",v8}{μ\>ڻ#QuKc Gpܲ?Ã7Z9W\)mU2(e*pbk{$rN D8RETÈĵAQ_bGu%oGQ~?g(CF-MN-Bu߸q#orI1"e>EvA$Aa]Jn vWJR$("/_e~8}"ƒ)Ssd2˯6SZ^+n\ 曓F[Z@<Tť?M347A'w֤Z>Yc'l,<J)aeR#`^ 3pqD)Po";ǖJJ#G JӞ]48'"U#2 !"ql" ;; OLtzBnShv `۠i8e. ʂ&D>&˹0YzuZ=yGN BZ‚qlec?""dן BaٚAB!$AF_ oǍmQɯ "& "gUD58XXD8ݢV !T_b|X,pMBs3հ^6,QiNJ4boܸM|-)?'FFymL"B4N}7aL?Dxv>NBrF"BsjtzG7/pNgb4꺥y+Z v@R8/W*wsOV9k'qRDR:6t/:6="#,U޲(4ϋjjdkK)~򓟴<3gT qT>S.jajfFK$J|N2(PN6"TjvPJ)혍 T(ɵ BUUIM؝^ETlm0$) 7I>>\.7۩Y_ʯ,|/p8+EҀRC=tF Л#y'p`ú*障 (ciy>"L3l1J#ƘI!%Fe]o^ {ڟYl u9t|Tsqr"ء oo} jœhB@@^+Je~1wj{C'j`a.GYIk YZvTdI-FL[qv72FYTLL=SGCE908QbRw!I zǷѶ(2DH_%ۏcD^_0YT^̩Ůռpnm鲜ab#pmȖM>+Tܣ, fJ$Sۍ=7C_5 "xĉ0#IMe1ሰc6Z)կ0TG{g5>"DDn$tGo#7 ߒGhqE^BA "B7prx=;9؊8"/-Wwc[ͺ&i`QUA@HG\V& KlgE}ɘRվ`r|+cieeA"Ed!,K̶dL5 4矻3Weh? M6(e0fYH*s<Ї -f՛a#$p]/}/' GVXo=/RD82rR~W`En蹾"4y$i*BxdfEl"r(0(ogbQT˅U2dtJ[3H58 4߉: r+ZjqĠpw #mn6 K|SqQt㻎ij4 R \w$LYFeg9ZRB\Ҋ` 2 2Kq= h$ϲq8F w;qjV+|C-ؾ^xM-kf<CP59΋mY"X2H_#  oc_PQL1AnT|kD<'qM]?b=K6iD""K.rXZg`)첣On~?؞vUiYe GYCbiY(!0uu, eݰXK~0x~ 3v:nmQץC%čC;~aQN ͢ ,Ҵ^L77kTL"t,24ZrrB8o;/в}\n_OҒ "d߇%dh^؞95@@g~?_Q:ZGO?MS/L |QJiN-G78/ڦ>h8DjPCMo&ah "n8'4YnjoٸǯotD8 |"Cj#Ys6~DXO@QJ/\L)?wjRj(H eD|o5j9>yR@$IFl0(ahiqK v~Smbw&noSpmuQDyJ[om<'t]YZ-P(gPJzQ_9Rj"mbM ׷ K.ݸ6biFH#Dn~xs_ʯ}ckN6}/zzn`K8Bf󽱍v|-3QrZc6R z2ۃ >9z3闟?G,w}`ow&xcRnvv|lz=V ?#Z\gmYh#"Z$feF-y?7Ak$klB_EBEA]OчVbZM]1/,_dIP,opg7>7aAS a/Mˎl dkÏjy#5~2'S r!!IYDgB](d尌*'58[s9MQJLaƀ:=RJ2V,y Tr5xli#7/|fCd=j(ZQPX(wG>;t "LTJ,"7dVA?Ep}tww׭])׿݉67O!8Ƃ^\("=G@D`#" bY 8V63JM)5K9:7M2xJ) bQQ4'/p?9]I 4Uҭ"Wj|(E8GB:58( $q<^"4yQॡMms{bz] EAPչw.XU;fR bs¡vysȽn~J{W jѽ=u;PhGHf>e5NXmz j_P}eȬi04ɢ{8`O+n/77i>,o'5 uJ)} C9]Dr4>ЌeW٧ϴRJ^LҾZXk vxID˂K(eƏ?]-ς yYPD(IP+T9!L5z?od( Q6dЋ[ݹ_<094@1Cxl# BtL5~) BMۇp_EfME!K7yG~Hs]-llTNoO~&IP*h׊6p,>mFiQ,EQDUOq"V*m)z0iQFoٸOR4gھN "K<4u! -l)"9V7òhQ\:^={BKm0ki@ {!tWBj#DD]9('v&3"& YaKXBEF5g>6ܡu9Ǣ$=߉%7d_Go޸ry?t{o=׽ ؀b<{!Q w/QERIN|)2hGqw;bOK.T qXylB|W,%F3NyPsǗ>um( "6\|׃|; No"bO;Ε[OZ6܂!lo"%5[#b.n2rTIq]Gyg(GdΤA$y[ux~_p * T ugTn84 motvo?Fw.|{&CH#9"̮%9~/xZ!"iE[oN H8bu ")ZX)"@˂HPQDMôUe= M]"i! Lkv'ʛjt/&lI!J2;fl,s쯽CZee#/}#Z :$߿QC_61`T>TH5U0D;ʎFQ1uM6ʹvҼ |R <[ith,ۣ"xtqf#e0Ԃ(plNxK(b3H Ւ0@5FW?.C`.)q([ 8+g}8p,2_]R; h= r[ lQo4hg 3N_`rTIya9 Sf"2gR],",(~@ؽ*K\R쪲,2o' %+9jyTQEMo`(6r/~ؽ?pX,Bq(3COx~ziӸבkI9T$*lJ%X[ÝmwrjU_;tz{9;[m"PJK%LQ%pדnywr"bZoO! R'5b8P۽=HλJO<DgkEBxYxl{=Pn15⩍q;gQDZX;}Ǻ&3D*9/5ID(O f#vp~EA$Bx3!^3l0gty{a=BI5Լmcp8?Nx;ID;c){ր@1gp/]zsw܂S0n[ҍνCBe8AUܲ  `ҧ_b9eԪn}`0sv1(^:=@,'ɸ"a0p BD4Ԃ)aD扇 <)ǁ'G5^2G<4GP _>TK5R 1^9ݷ7W0""Xx}$Y(3{AA0i#S?,M7JVVq=̖ W,?rbCY*;{u03 @=WJf:e-׷NdY՚-HN)t\Μ/~q,bg+!@R]y^3jؾ[DZN-{P.L jqj"/6z䰅0PJK%pEi O>(eB&EN4rEjwH0*ZtuW' g*ZXN9M`cN ",/'PC{:[YBe (UY}Cp+NOTozmL BB0Cm0 %โF3eJjRʊ5Vr+$ NQj\XS%0|^D-o $Ȳ1A7/|{w#밲t̙6B8t[ޓloFT5jkNZ ; IL0 a12҃# P@z#VA}Ib-mcz TZyVRU]Q ,@,sBAp]P:o>W_su&vAg(KJ=~ ήW} "8 !ٽ sEDߧ&0AP,"?͡Tv,n GfyrB9!в!B/pOdP ""&TʂCl!<j^osgq~3+eq"6B0 4^׃R e Lkﭷ߇bqqs؅oYКD% x^D86œZ܏G8LwbrF !&;h55( %d[ª %c5;tTzW˹eYPzow\ۣ7`~"/Kl{s+ufzbuaϞf+tΰΤcDX,"qsJz!yNPD<{^Q˜QHB 1|>qw-;1ͯV̳Tpw~ZV S%5̶l6)a3vikr"Ta t0)4@}Ge$Lg&sal*+_ILw}~ \3nӥ%؀xz,8C^9UW~Mozӽ;}>vmwyLG gmq JonN-C\rd߲~m`ceHD<P`;_V$b{V|uiZ QN i\&7K4zuW'꽭88W[oZ33u"ht:W~hYP*vdU%ԊĚr ^4;i#A@i$00-¡p@UE1E g!\9QvF˅pЧ\-@Ģ^@g0dxIDA#8 Jcil]Y58i CZFb+!ps"B "CpMuE0HX"OX s尴2Zmn)hne,kyMo>}/(ٖ&u`%\<ÀZUnCTU1aDl/;j=G6 `4 $ gܤz?6?x?#`%'߽ @KKq_u}-oʡbD78S?G?wFLZ@ 'k#D'pw\˅jaلr9;:[[4&X.U9#mۨ54`?"m焮dGE)`,r~}"Fkt zowmCjPo%n,kn9R]H;fcZ 8O;<e鶗N3E!%~9hÈDqH" ,DKYFIRAB&ҤT⋶- h YmI{o7[/d[6)\ZWpJVɧa-mł0oFEDET@DUE^ fv̺[[ԶJk5n;~Dj;QEgHt.CV"N"C_ňg~gN6=j__4ǔeE%[9QC.u^$Fyj5dGJjō( f}/pcdRA\Y}o)-pXL= BscNE v|K{"k=|a+ܸ! 0_)!BXZv{44"^/q9ތ7V4AAׁێͩppTHdF6,b~8e.% ˅a;Y-^z2?͝q>*_{"\ QvA-ϢdOuYT۴PzW,2$(!7BɓҠ2D x^,~Vzbo\TTe7 Zn='AF|O+W>0NW`yIDeqlCzefʾ.|JPDC"_!"t)6;6nw/4nay)<_Yl[dvrN>ON8qUsy00"P"(Ձr dI^ q}=ANmi(^iώ&wtEE~|GcE^++4WշͿI&U !AXR,bj C!q^o}ڝY!sYm?4qN7Mc`eۣ?<7+4ڷ{ CIљ0 $7=3~!lTrY.!<ZJq,W{f+dvV ze`w X[jni+_=woZ&x;DV֏;fg}qgj5 '#輳8rN}9=DҍΓct߯a$ g|mr@pcC~?B'W,5Z'ʕ\6x 2Bo&s0pla"4B|Gݞ>oKze?mcET'}NK<~zNwLoO6T;ށ<$5Mur J_ňpzEX"b x@@$ D K)elԊٖV9xC\PK88"d|4( ]P Zs*B'e#ҵ 8z@[ӣ2F !Ge=qre"BD8}/^ܿ'@H=W+8ب*imzltd}wZIz{jax(F++FJn(&X0 ޹;ۋYd^waQDYΦsxxqh47-Ù3\qxsnL5:Ge] k ,/c=]2$AoK(ju?!dhv}2n UEn5 8#"9TٙJ{O<Ƿ ze*-`c#Ar<~+ +r}7@snr/an4B;;;?~{o-4 P"B7]` ~h#:Z@mvɝ3[ 3<nXs~yqaĕE-{4Ev" 9"F^?2Z:1i˅(KO^kaVE>J3=y776@?Fp4\a "Dߓ,o{R9nR~ BLM3YYU6 %p~rЕ<+\]~GOw,hm8ۣ]SšO<Ƿim ,qע~u|xN '= 4c "xPΝ~s'csaC,GԛyV'+FrvKw~۟_Gֳ;ckAlPRb˶B!'MD,].)KrjNշ۲1ڂ%8v i%W :JisR:V>q~i8x6|InM4*Fm[E),qx׻z zx=e- U۝LiF7tm Ck ]ZίNM_}E/n yr糆WZWWF!/כ^}_}=kRIIx_Fk&WvG5E&[[qϾV% N\mcS/\}ߎ n/+r/< OO|{~ԩSƵ0ը<_sqrg[&q6JVHE $dd@#;sc֙ g_F,"ygvvrf{G3#$=g~}5izpG>JuZ[,LBckBdE!Qv d7Q% t: s >#M)I("q<Aa<)8@ ID1RJux]+Pq{Dp ǃUi0pє\٠0>b`2 ZY!$C a~(+±@r0"sīODO镩,"*b0'x>)ҥ%TBi}>bA+yTdk8*GvC2 ǣ>p J+B&*@ssji_+@7L?4}Uo w?p(&tD#$@D"d!̃snu˭a`D#N' 2QqJ Ev].< (U@diɢnX> n4y9HL[Oq`1I]nL6D#1ȶg6H090op@%04!ꂲ2d2!b hu25'Sں< qSЅ/w7;_|Q!ۿ#XБ#t!eі#nP$au+FQ!Tiou[~^Odk4=B2I& >!adׇʡڱDEhz_iQ #D~@ɱ? 8a̫ȉb}ŦGwNٌ3b4JTv^sdrAi)x< [1T(:ЯF#Tƥ$(` ٟ8HDcQ#~0 `CKq #jѱ^IA8LV`24W-La $D,"AB(S'x2&0ۋf(+C{AGMQ=%$غȿLVL^$Z !P(`BOޒdS@*%GlRH{ExhF *+ezdx1@@l9  Od6#HUNԄ P _Ēf3jlSwR;ۋy!"2 Dʴo`(y]"nAbq`^e0`ti >\V0&Iz7VWz%:<"]y'2q̘A= Owa!˲A0L !$˕Je*X$J3Fnh#z=rqͮ_iksɖGbDTΊӁ}(6.bRF^Ǭ)I I!`Y‰Le2PȩdTt1eBN6wX,DZEc)*3.S;bG1HSVHVBR::QN0U."H A*WJ%Ǥ Rq(ˡ6UHZhvZdȾk"X">OFW{DL8,rJXbЇ"doPBe`xP@r@00VrMxF|4ɧ1S>H? }W k*1^BVU%[Zhc0ZѮ.P*sۘBfs)4'RiBd!q*ee5(Bh+ ߾.J%b%Zw8^0ǂ,Z,I$B,Q(U I?pv~R!1zo5Ǣ)\G'8u8W] &l:QL?sﰍcl 'X|\Ƙ~F8BkȢDDb{zlR6(_$NऌlVaIDZDt0aC1 ͦQ5qXP2&˙Q@dLd CKF< CâXbz*}l8 /7Xx'ݘ0D*2 v|w;r>R"D\^X8Hd /BPXw@H玿rˑ}aNvgxʈG4őxhlp}"]9JhBbNNK9 sP\ &B@L`* l <Ė1E~ZXqhD_nՠ6qU,&"r8 YzYZn>~dѠlhḁ 8@:CSocԄ VP"Ak{#G:DyףdT 8ҏ bH wB.N$pz b#TH8sGyC:ΘZ,`2A$osz*R˟"{-R)ZE)T \B߻̩3Od uH'eo)FD< l&@K$A8 Id4A4;3PBHʴC[bTP`{:G1Df= *7v85˜\Z$4vO#AfYQ 8±@yT 2=HC6J(dfK'LYoԿ"̈P*! %af!S6ɨ.AY~\qRydi%!U8p(U„rJJRy<4s25Ԗ`GX[HD!A9ͰlYjpt4$sp$.$d#n@qYupujp<ͽOhl 2p,eCR{$W-<}ru4${mGu\F'D25Bv 9"eL :EyRH#~ҘHW.-@DRij% -Q+tĤY8f3gR)~( ^,R !+K IrVɵ x$iBCѪS0nΐ KeF͎j;8(5b0C $HiqJ5Bɘp% ̛ wLVR r]HG]Ч۽Dn#AQH9 >FH8 ` f~DR= 2BYF%`YɆu.x͆p}n>_֦aeZޮ q_nVch0Ä! 냢"AʠF!A1vlIsAz+󼬇V!ɉP8?d~ ^5 ۚ\,_-h:1ǁQS7-NExYC 0dXjYi ^8C )`[$Z*S#\w &Sq4X}!w<%+B)9l`ͅL؟JE!)KloYSLQ бþ}$(?% c(h'߮6Q3 d ׇe2&=Lr9c[p+F1v*0P=xC, P*mP65H41Fٳ$&u1jiɬctZ%TZ ,FR%('+^/H$Ĺ4DZ6rav>FH~ II"P(P!8>GDfvI?bm""jGL8]s鍯\:pMMDmT5P62 {,F|& OFSBB(Ƅ5!d0ƙD*%t ch#~ +0Z$S'Tj@F '^J!R"M31T 2|u:@9 h5Lk'~19E"XNi"!cd2N0 L̝m@[4 E qxB#!4cu@j!#[,(6BZcLHa@At֌xС!ۍZPUp׮c~Jb8;X WZ|!;/|"g z,xR *:y\ID>-1Ʊxe$idjlfYR'r"@.Q濟caJ mgswJhq({x beR (7˅=Z-+H6ncˆRZJ<>!`&aAZpP#ġd@QHMݎ zꩧfΗ W# 0 '>[;z+S~q_p|*_B5?c-<\B"ƅ)˥ [?LgS 0,IZs#*($cߟ)MiK4 Áf*acڅ?4 Q>G#(y|tUi3rzP@43XR @aW~Sdjb3eeamVw_ۢ-/V j#$|{ 'OL+ss|J !22O&Y[1A.tCfy)KX} :昍WRQSDuDca^5բDTmՕsEn7` `kDq6ӨgĀt4bVF c|}ۑ]WNe !B`_S&WnթMu V9HbT$s2UH+jNBؽy] X!GKYmKþX1N'X7S+r,pX"Pt:&f?&a8~A!hjg q!D:=Nd# RVKFnxPBRc8v /\HE  ٫}qݍ#~^N(/GvCeO=]d-ؽ0Sϼ:zqBp#yk841Dv}-gpƁ@G"87nV +$e Hl#HՅɖ()7ʵm"$0mVoތˁ$(pԃcƜ$^ϰ*egd"(+*cfC==H"P( ˂LJ}c aLDqB -JKY_L`єm`CQpF D!UԯLEF{{qq1`*\4% *7w2JD}`R)R*-~ …czEML!IYH,fIb3e4Y C!lFr2'8ѷml$0]R)z1cS~/ţ &;YZ}\[M@̄@:;qi)Bѫz=$υ3TC; " шJ\׶e-h9ֻgu+F^$8&D Bm-ÇSG2V!FVW#/nD/ŜLmڗؽ+hzRjTeb?@Ujccxm]fDldTctu%`I? 2[Zk,7Hm8MIzm먒 `HkGb&%˰|F,㮮h 8GDlKd2 5cF.A;7sP.`VbH<*|6BR# cp@(-gh3 jK3G6AȦ?+3yMQñ0'J7ǎaz7`{ޚR,XB--f v;TZ??>O'[[ .OMB9xԌF$cRpKnpt: ٚĀBzTOM܎ ?h#laFIQۇ l!DrRj9c"02 #-.F$!TdKP\Bà!wK9T9le2&B2)'e#D XXjz)+2O|@FlRmj500E7lIQ`2Ar9%h,w<-g"x,P(c?!d0hHu{j#ap6لG_p$ϯ&Ӄˇ$d.¨&E[&JF]n3Uz_d>qrFKD}CP ebjna`X ` evk55V\UL&(-{s'V^pdjNN$k0\ЀkOٌ@Uiq5 (] G+.DǙz)H\? #^a0v,2=㈥Jx$;Rӆa((+C1D@ ]VH`} eGŌ%[VH`CQEc,e x׫Z#hҘE>w[7 lēx2iNr>u M={eErb&0b&R4CS,0̉i"ϒc| TT I|=6B܃z1Ơ4BȗUh,ҏwKIX12I,jk_yŚ`0 UhKD kj8ISܛr;qrFyޠr|> PSlm% goM jn>N҂kjEĉh.\]q- v8FJBF?1bcQw7xCmv} j 2LZуRkL0,ahvZ<QP<'Sm2r*ҕz: EBh<5jD z^`( p . MRVWluRdXFr?_:_&]Me4^e޿W"Bk Rj8O: b1()9؜2#Dei9z54(tvQW#hdjnCTX aj(+C۝nZ[1)I2;;S4'Gq8OAeai8S Y2 dt s2nŲuwc'!*WD+$+XX V0,JzK8x^^~@)+ i4P@_`YF*a~{+Z# &"hO6*mޝ͟a#  $^O=qShP#%"B-D CE$'TQ,J`I$]])vϒ~aʔԝKa$L%Í7R BJe{ :ƘB+m<;7O3lcuJSgKь qh w,{ܱe-5arbHQ ~MqC5 rgZ^JDקDZDdU 9cJK-~u˦xǍCqĝ;V㛑O↮f4qZatӁ74K U AnO$tx!BȢDCryaJJa2 --9HYnU%Y!>_p"lv'T*8-hD?ů@n7\LŲ"hBwƍ(NorRkr5BH0BhD# >x*Պ| D|- fooKD\N:JgBIҤ)JO]/&D$Uh B$F4aFbCw[%azK)BǁҨ4W9 騦3+ Ɇ~$!9m {JL1c[Z0 jL8KK.B¤L Lej r2#*9mt:H&!VL2رq)& Ғ~lm&SjeawofOpTEv8xLD{E;tJLUjoJ$ȣb*5PT3_;I9Ch m*ҕxv~+6t}fmNҠakcQR &hLRVNQ4W#h@.;S!vA \24Rۥk,r c޿?2ZdJ7Lhƽ5NduNSioo)m2G^o*(Y>W QՖ1-}O~k70 RMMvH>f zܙ`BرԷ|K *<ލ@駅ѬY1aMIjo}}l6@uHMyt))$BR:!˪cUyPF xT !Dr<BC[qwf w ?{ Z0>S^\ Ɏ[E%%pfTBR*b:#8|o+VP 2j:!`NJ/+=m֭djW\ ~rRL)RAooj_}[*/GC3t=_ȭ:hzR #XǕO%FW{=>?H$H0\)c?g<9X9E0YNPSir֩-8F򇒆)$,Ƭ-͐@s3 Qqf;pz TWe5e $?A a\_uu0469)zW#%LS4c٦:]w9xBls|4]-,' .|Qyo&DL\Ne(b˃ٳQe%zcp:!oA`}޽ˎ땔Qt5g7Z 3^-„ .]h[ټY3eg4 |\^/|oT`rBr.RVFS41TCG҂G:iB!LsPG"ۋR0W^n"`YoަI7iD2v.m)EMp onm>\DXm^)lLr]__{5WwIZ  P_wkZfCMM B@BBƲ}&W^RK ph:A\ᜟ&j5zZ"/g_[.s 2,.̅ a現?^n%\ʧrѓN\@} }7" 9M F4 Fm݊w:;qI #,C07cS"$W:)]([n ;f vp=|$-}_ssՅ6En'RS m91))ZAg{zt([~ B4i-~:ߏs {}åD*鄁r'qЛo SSG$KPf@ˌd]tM$RG.QM%fPBHRlIƍC}}rAk+TT Yub W6O T*G?ylFIJ0 N 144 /YBeE;1Q{LXn,qB־< ͻ81Vdg[KSZ]pn,#[' ~szjg瞣ozec!W)!R/RVN!I#2t焖!B`6`/\2V섷Ӧ{索 ?cQE1_qغB]NiȦ@5ٜ޸:[o rH4¼A _*EK( 2cέ--OqIq0ƵH*Eyr(.L`dRj>jjB?IA0mZO xj5KԴG}մieFCCV+h4؈3CKB;flLf\zj]EBÞ˲2DP=^-,[F]ye;#/I=1Kg9I8Nw5f1N|J)eaBE uǩєACp]5P[ᅬ-;etydexR̜yb#D#ܲE8\dP|BMcx]A"7 &I(aw7߅?PSӦQ8I/K ܜJkI}!)cFGM衇JryAGOF,h{sh9Q~=%Kۓx<ǩ UN(]]xF.hno`en{%#B.+؉B`̘AgorU 'g lꛆR ho[oEz !rrPѣmތag<?q"=[t9sЌv_|!ʏ3p V !PVZtY:WA(/GwA]}55zMI>n :hY^!5*"N|64ݻqg簞2k\k=8Zg\f3('\$!*7R2ʕԒ%d$sWG>JكyO$ ^NEƝ|7O&G{?}T*˘G_TxsAOMpBA$mӮ&4grDQqPWr KKJBp>zB! 9a6M ޚX2 F^K=t}71 a|ce᪫N(T?Ɓ9ImTef*?Q?(#8. *O4 Q8|GP};O<~G3̩жЯ~E?? ӧS'̧=s@XA}bTQMܔ(-HؓZhէS@ !:IBơݻq$"qNL pT:ľs^ihw݂ ~9a6C}qQbP^~^( jtčߒ2]UeFZ~5TV:V+`t9gfHcc9jX\vulEɓǩZ ѣ.D$[Ʉ_Gj MMMv/2׃?5eԼ@1q"zaj3k N|UVm޼y&iŊ<ϟh :%`6^r9%@M0rOf0)E*g~hov LH:3GCW^-/A4oh/˷݋L<(=}zjW;X 551D"EEp\y%N$ ?œdw*ʯ{XxzG`uN%xO$cfҒ6UOxDW1˞pH&Lg ԙQQF555z3gwmX^|'475A['Nmw[ZN&Hr ôiӦMv,,,d?݇|H&v]Ӊ`۶m6 At֮][ZZZ__ܼjժ{W7H%wǍFhv\PHPh4t6qgqjF`H~mmmVU܁=z`08m۶;vL(cXwwwoooz;)Hhv\,.].׶mU*w>r4>6mڄҧ {<'Hofs ?zEN=cÇ+a4S{D6mnX ۵k7{8Hk4UՊ3m޼l6_fssFpꌭTh O9%KTWW,K<_0ޘ1cnE3G*//Z+Wӟdt:O?}Z,+***+++**9s\{˖-;wn0$򎎎iӦ,_?f۲e XO0eʔ{BѼD[WWlٲolƍ4~gD;3z{g̘1({_y .['O|x<1nmm5jԭ:s{G<wF+++l61ްa%\rm]z饓'Ov: B49sҥK.]:{h41~',YrwyK.bG}bŊΙ3g?l6ێ;0ƉDK/]d?|q| &b˗WWWK$gtࢋ.2 ntL_nw0,))BC AX|c=F䍍s|7۷dq!x<&,GD ƍD`$YjUB 7pv!*+++;z(in>tPBHc֬Y+~<<`^x9F1;~+z.o9sL^lٓO>IyWcxdEEEBO|`0fqIT*–-[͛R{͛  ڵk̙9ǰ8>;;;G}:/|DlڴiܹEQ5w/w^Zm27o<}tq?3{lr͛Gu]~j,w}o?LhFW}ݻwtk׮%ه0 ?ꫯ>&wygɒ%98 ```0F8 644ѣ>!e͚5}GGի \.BP(ȿz`^qb5A|s@ `4rPL&2 w H]w^lc1`͚5Lovhoo+} i?3Zwtҳeeeh---moogpt??|7c90d@ X__1O̙\!af͚wr}"cՑ)bY~ԨQGyw]xx 0jzѢE{:#G;S5nW_} 7nF ]f6~i4ݲe֭[\s͊+Nƍ= ׯ_p!9{ƍNß|I4---EYp EV"7ndΝm۶m޼9}g 6dC.\`Z?۷K7,)H6˲W^y?_/6"}'x;ܹsƍB+V+x MZ[[W_}5yqu׽۶m۶m[o|rxWw/߿kiiQ*ݻaÆ۷袋`ҥ|gϞ^zoڴi`Qr# ۶mۺuG}h>Cb<=8m6mڶm… ?#z{ 455yǓkijjZfMMMn'Lׯ_jժf ì]]f L5kyfm͚5~͚5ƍ;Mx:oz߾}wjA@dYYYJh4&%%l+W{* @iiiNN,Y^^766pj5ӧO|>_G<uJ0gǏK,Lp\6h,X]yJUUUpÇ=HxҺuܹ%I]]<m;BVrrr]]]DDtMMMw0  X,Rn2 hܸqcCCÙSSSW^-Hlɇ'͛7J%r\& fgpjuBB|*+ Z-㽽yyyA {zv'IRѨT2 zzz^|aÆ"l"ȴ /^,(JNNvlE1}y"ƍ 4/,,lG~:ɜ"˭Vkvv661mxxX*I<<K /Sb v;̴SRRRrss1 KJJMEr6 3}|Џ8 ‹/|8͛/]{??aNh"@PZZ8F!jd2Y|||FFInwccH$hL&jRvGGG].Jw~;;;VdUWWL&QYY ˻nxQommAAAjjz{{K$Bt:t`0?(#ȯCih]v PVZcx֬_$L$Y,T*rѣGp_|yyyyjaaa``3g> endobj 374 0 obj << /Length 11160 >> stream q 15.000 29.575 577.500 747.425 re W n 0.965 0.965 0.965 rg 26.250 638.759 555.000 138.241 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 638.759 m 581.250 638.759 l 581.250 639.509 l 26.250 639.509 l f q 35.250 650.009 537.000 126.991 re W n 0.271 0.267 0.267 rg BT 35.250 767.476 Td /F4 9.8 Tf [(Fig. 1: Nowcasting Errors)] TJ ET BT 35.250 748.106 Td /F1 9.8 Tf [(Percent error for three years worth of nowcasts \(forecasts at k=0\) using two models: the baseline autoregressive model )] TJ ET BT 35.250 734.370 Td /F1 9.8 Tf [(that uses the previous three weeks of available ILI data \(green\), and the improved model that adds the Twitter estimate of )] TJ ET BT 35.250 720.634 Td /F1 9.8 Tf [(the current week in addition to the three weeks of ILI values \(blue\). The vertical lines mark the beginning of a new season. )] TJ ET BT 35.250 706.897 Td /F1 9.8 Tf [(Each seasons estimates are based on models trained on the remaining two seasons. The model that includes Twitter data )] TJ ET BT 35.250 693.161 Td /F1 9.8 Tf [(produced better forecasts for 86 out of the 114 weeks shown in the figure.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 358.632 555.000 272.627 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 631.259 m 581.250 631.259 l 581.250 630.509 l 26.250 630.509 l f 26.250 358.632 m 581.250 358.632 l 581.250 359.382 l 26.250 359.382 l f q 450.000 0 0 192.750 35.250 428.759 cm /I54 Do Q q 35.250 369.882 537.000 52.877 re W n 0.271 0.267 0.267 rg BT 35.250 413.235 Td /F4 9.8 Tf [(Fig. 2: Nowcasting Predictions)] TJ ET BT 35.250 393.865 Td /F1 9.8 Tf [(Nowcast predictions for three seasons using two models: the baseline autoregressive model \(green\), and the improved )] TJ ET BT 35.250 380.129 Td /F1 9.8 Tf [(model that includes Twitter \(blue\). The ground truth ILI values are shown in black.)] TJ ET Q BT 26.250 324.410 Td /F4 12.0 Tf [(Discussion)] TJ ET BT 26.250 304.456 Td /F1 9.8 Tf [(Prior work on influenza surveillance using Twitter has not compared results against a simple autoregressive model based on )] TJ ET BT 26.250 292.551 Td /F1 9.8 Tf [(ILINet data. Indeed, we have shown that Twitter data alone are less informative than this baseline model. Our work is therefore )] TJ ET BT 26.250 280.647 Td /F1 9.8 Tf [(novel in that we have established a baseline for comparison against which other models may be tested. We therefore )] TJ ET BT 26.250 268.742 Td /F1 9.8 Tf [(recommend that future influenza surveillance and forecasting methods compare to this simple baseline.)] TJ ET BT 26.250 249.337 Td /F1 9.8 Tf [(Our analysis is the first to systematically characterize the limitations of ILINet data. In particular, we have found that forecasting )] TJ ET BT 26.250 237.432 Td /F1 9.8 Tf [(studies that use historical ILINet data must account for the fact that these data are often initially inaccurate and undergo )] TJ ET BT 26.250 225.528 Td /F1 9.8 Tf [(frequent revision, effectively increasing the lag between data collection and the time that accurate numbers are available to )] TJ ET BT 26.250 213.623 Td /F1 9.8 Tf [(health professionals. While others have noted the existence of revisions to ILINet, )] TJ ET 0.267 0.267 0.267 rg BT 380.672 215.130 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 390.310 213.623 Td /F1 9.8 Tf [( we have shown that these initial )] TJ ET BT 26.250 201.718 Td /F1 9.8 Tf [(measurement errors translate to errors in forecasting. It is here that Twitter and other social media data, which record signals of )] TJ ET BT 26.250 189.813 Td /F1 9.8 Tf [(influenza prevalence in real time, can make the biggest contribution.)] TJ ET Q q 15.000 29.575 577.500 747.425 re W n 0.965 0.965 0.965 rg 26.250 638.759 555.000 138.241 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 638.759 m 581.250 638.759 l 581.250 639.509 l 26.250 639.509 l f q 35.250 650.009 537.000 126.991 re W n 0.271 0.267 0.267 rg BT 35.250 767.476 Td /F4 9.8 Tf [(Fig. 1: Nowcasting Errors)] TJ ET BT 35.250 748.106 Td /F1 9.8 Tf [(Percent error for three years worth of nowcasts \(forecasts at k=0\) using two models: the baseline autoregressive model )] TJ ET BT 35.250 734.370 Td /F1 9.8 Tf [(that uses the previous three weeks of available ILI data \(green\), and the improved model that adds the Twitter estimate of )] TJ ET BT 35.250 720.634 Td /F1 9.8 Tf [(the current week in addition to the three weeks of ILI values \(blue\). The vertical lines mark the beginning of a new season. )] TJ ET BT 35.250 706.897 Td /F1 9.8 Tf [(Each seasons estimates are based on models trained on the remaining two seasons. The model that includes Twitter data )] TJ ET BT 35.250 693.161 Td /F1 9.8 Tf [(produced better forecasts for 86 out of the 114 weeks shown in the figure.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 358.632 555.000 272.627 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 631.259 m 581.250 631.259 l 581.250 630.509 l 26.250 630.509 l f 26.250 358.632 m 581.250 358.632 l 581.250 359.382 l 26.250 359.382 l f q 450.000 0 0 192.750 35.250 428.759 cm /I56 Do Q q 35.250 369.882 537.000 52.877 re W n 0.271 0.267 0.267 rg BT 35.250 413.235 Td /F4 9.8 Tf [(Fig. 2: Nowcasting Predictions)] TJ ET BT 35.250 393.865 Td /F1 9.8 Tf [(Nowcast predictions for three seasons using two models: the baseline autoregressive model \(green\), and the improved )] TJ ET BT 35.250 380.129 Td /F1 9.8 Tf [(model that includes Twitter \(blue\). The ground truth ILI values are shown in black.)] TJ ET Q BT 26.250 324.410 Td /F4 12.0 Tf [(Discussion)] TJ ET BT 26.250 304.456 Td /F1 9.8 Tf [(Prior work on influenza surveillance using Twitter has not compared results against a simple autoregressive model based on )] TJ ET BT 26.250 292.551 Td /F1 9.8 Tf [(ILINet data. Indeed, we have shown that Twitter data alone are less informative than this baseline model. Our work is therefore )] TJ ET BT 26.250 280.647 Td /F1 9.8 Tf [(novel in that we have established a baseline for comparison against which other models may be tested. We therefore )] TJ ET BT 26.250 268.742 Td /F1 9.8 Tf [(recommend that future influenza surveillance and forecasting methods compare to this simple baseline.)] TJ ET BT 26.250 249.337 Td /F1 9.8 Tf [(Our analysis is the first to systematically characterize the limitations of ILINet data. In particular, we have found that forecasting )] TJ ET BT 26.250 237.432 Td /F1 9.8 Tf [(studies that use historical ILINet data must account for the fact that these data are often initially inaccurate and undergo )] TJ ET BT 26.250 225.528 Td /F1 9.8 Tf [(frequent revision, effectively increasing the lag between data collection and the time that accurate numbers are available to )] TJ ET BT 26.250 213.623 Td /F1 9.8 Tf [(health professionals. While others have noted the existence of revisions to ILINet, )] TJ ET 0.267 0.267 0.267 rg BT 380.672 215.130 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 390.310 213.623 Td /F1 9.8 Tf [( we have shown that these initial )] TJ ET BT 26.250 201.718 Td /F1 9.8 Tf [(measurement errors translate to errors in forecasting. It is here that Twitter and other social media data, which record signals of )] TJ ET BT 26.250 189.813 Td /F1 9.8 Tf [(influenza prevalence in real time, can make the biggest contribution.)] TJ ET Q q 15.000 29.575 577.500 747.425 re W n 0.965 0.965 0.965 rg 26.250 638.759 555.000 138.241 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 638.759 m 581.250 638.759 l 581.250 639.509 l 26.250 639.509 l f q 35.250 650.009 537.000 126.991 re W n 0.271 0.267 0.267 rg BT 35.250 767.476 Td /F4 9.8 Tf [(Fig. 1: Nowcasting Errors)] TJ ET BT 35.250 748.106 Td /F1 9.8 Tf [(Percent error for three years worth of nowcasts \(forecasts at k=0\) using two models: the baseline autoregressive model )] TJ ET BT 35.250 734.370 Td /F1 9.8 Tf [(that uses the previous three weeks of available ILI data \(green\), and the improved model that adds the Twitter estimate of )] TJ ET BT 35.250 720.634 Td /F1 9.8 Tf [(the current week in addition to the three weeks of ILI values \(blue\). The vertical lines mark the beginning of a new season. )] TJ ET BT 35.250 706.897 Td /F1 9.8 Tf [(Each seasons estimates are based on models trained on the remaining two seasons. The model that includes Twitter data )] TJ ET BT 35.250 693.161 Td /F1 9.8 Tf [(produced better forecasts for 86 out of the 114 weeks shown in the figure.)] TJ ET Q 0.965 0.965 0.965 rg 26.250 358.632 555.000 272.627 re f 0.267 0.267 0.267 rg 0.267 0.267 0.267 RG 26.250 631.259 m 581.250 631.259 l 581.250 630.509 l 26.250 630.509 l f 26.250 358.632 m 581.250 358.632 l 581.250 359.382 l 26.250 359.382 l f q 450.000 0 0 192.750 35.250 428.759 cm /I58 Do Q q 35.250 369.882 537.000 52.877 re W n 0.271 0.267 0.267 rg BT 35.250 413.235 Td /F4 9.8 Tf [(Fig. 2: Nowcasting Predictions)] TJ ET BT 35.250 393.865 Td /F1 9.8 Tf [(Nowcast predictions for three seasons using two models: the baseline autoregressive model \(green\), and the improved )] TJ ET BT 35.250 380.129 Td /F1 9.8 Tf [(model that includes Twitter \(blue\). The ground truth ILI values are shown in black.)] TJ ET Q BT 26.250 324.410 Td /F4 12.0 Tf [(Discussion)] TJ ET BT 26.250 304.456 Td /F1 9.8 Tf [(Prior work on influenza surveillance using Twitter has not compared results against a simple autoregressive model based on )] TJ ET BT 26.250 292.551 Td /F1 9.8 Tf [(ILINet data. Indeed, we have shown that Twitter data alone are less informative than this baseline model. Our work is therefore )] TJ ET BT 26.250 280.647 Td /F1 9.8 Tf [(novel in that we have established a baseline for comparison against which other models may be tested. We therefore )] TJ ET BT 26.250 268.742 Td /F1 9.8 Tf [(recommend that future influenza surveillance and forecasting methods compare to this simple baseline.)] TJ ET BT 26.250 249.337 Td /F1 9.8 Tf [(Our analysis is the first to systematically characterize the limitations of ILINet data. In particular, we have found that forecasting )] TJ ET BT 26.250 237.432 Td /F1 9.8 Tf [(studies that use historical ILINet data must account for the fact that these data are often initially inaccurate and undergo )] TJ ET BT 26.250 225.528 Td /F1 9.8 Tf [(frequent revision, effectively increasing the lag between data collection and the time that accurate numbers are available to )] TJ ET BT 26.250 213.623 Td /F1 9.8 Tf [(health professionals. While others have noted the existence of revisions to ILINet, )] TJ ET 0.267 0.267 0.267 rg BT 380.672 215.130 Td /F4 8.7 Tf [(28)] TJ ET 0.271 0.267 0.267 rg BT 390.310 213.623 Td /F1 9.8 Tf [( we have shown that these initial )] TJ ET BT 26.250 201.718 Td /F1 9.8 Tf [(measurement errors translate to errors in forecasting. It is here that Twitter and other social media data, which record signals of )] TJ ET BT 26.250 189.813 Td /F1 9.8 Tf [(influenza prevalence in real time, can make the biggest contribution.)] TJ ET Q q 450.000 0 0 192.750 35.250 428.759 cm /I60 Do Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(6)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj 375 0 obj << /Type /Annot /Subtype /Link /A 376 0 R /Border [0 0 0] /H /I /Rect [ 35.2500 428.7592 485.2500 621.5092 ] >> endobj 376 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/files/2014/10/nowcast.png) >> endobj 377 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 378 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 377 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 42823>> stream xw\Ggw:DQ,HQFb4jby%c1D_[)jX[DEPAPQ(w\z11&;|vggi`A@ +${+@ =A@ dѮA@  ,@  !@ 3E@ v j"]ZhנYhא!$H"d@  E DBDBVWWfSR) hrARi%gϞ@Dɓ'NܲeˤI۶mYvm+$-'Lpqȑ#={tuum=M|>`۶m&M RcO ʕGf2fl6ZV FTTJUZbTTTSTLF"*//DBC$t:K. `6 _HHHuu͛7drppի|>AwƒdduLhhh Ph4l.--!!!~~~FQ?~8!!D"ݿ_*FDD[(++^^^aaa4MVL`R͛7T*Upp0D.ƲX,@eeWqq{HHN+**tH$z!,&l h4\.:vhU,+44>px䉏ϭ[\nTTUrөS'?BPє(.Ι3zjС]v&);;tѢE.\(..NOOw333322ϟڵnذ`L2%,,F*8qfO6!,,l'NP*Vp8_}Ո#QTTԻwu֙Lcj Ry,̟?q*zҥ;wSӍ9D">}۶m ^xqAAAtt}>}[.;;W^"h}IIIܹH$:x5/d[u6o)ziii!!!^^^ hF X,"رc<oΝ[lӧύ7Fm[ "---::Hҽ{j41c0LCOGK,ϟ8qbjj^ _o&%%ݻw/<<믿q{ɞ999o;s秦T.]̞=ū=#m~aȑ?H$+W3go%Kp7C ٽ{d _n]FFƼy,KǎE"ŋ Cu:R|W9( rj]]]@@gLL̽{p>v B (w0؆ zlFL&Rǎ5kƍgX={B%oݺաCJ84~O>W\/,,ѣdzUTT㬬O>qU\\x@@aI&m޼Ϝ9s֭2 񆆆 Fc}^oF,cvʹ4 lCQT Êl6aIII=RSN?~|VVf3l͞=͛-7ۘ|>_.h4Z\\z~aСC Ÿq:~x\^]]m-22o߾ ca {)//>|O>ҥKaGyѤe˖¾SLfϞ=alnnn!!!766QFm߾}+VʧcH"v$۷oVVO?%$$>}a>>>_~=B5jԾ}l:t0k֬4ww1cƔhѢ7xh4}wBКkqJ[nׯ0̙3> 44/T( .lhhR -իW_~˖-틉w$f"Ç7|>L&}ѦM(JBBBRR5~ZmM?СC4ĉu:… ׬Y3fg?~FWW۷VZ{ۗݻ6 T*_~$dܹfom 5I'bM,wڵ޺ Xq={g}fo?9ٳgȐ!>>>VᴬY&??`0teN@Eh4uyݻN@+`G&a0C+.ӧ[sss-Kbb"aR4==}ذaV ڜ:ujС᜜44رccƌh۷ogggwd25Qٳ*jΝ...O?{ȑx߂t:]UUӧ?~ g՚ܻwoذaݻwoղCj"|Q^^^w_ pD|&IĢEf̘N>=~ \޿R 7o90 .4Ldڴiׯ ۽{_~ uVfff@@@}}~D"FRw}w̙\.~Ɂo~MJh4*׮]UUU>>>hI@-BsioMyW6/ ///p8/S[[O< ;w.y}-[^#֭[G8qs#0l˖-t:O?]hѷ~;s̉'Ѥ_rǏ bYr%t{&;u$  4h+W\re̘1/_>BJdffԼ|xzVԥK3sѣGCChm)VVVFFF0 xɟEZ...XYY 11 ?d "2qr7{7nd:tСʨ( ***_ϯ{]]]m$;V9jUk愻::"K~'hG&LHKK5kP(Dzxb:ѣk׮__8>L[&O>>Ph/ 9 8rn̜9s9„h!~Z Zy-[wW.^F8'h@ "z^:d1%i@hf@ / 2D3#4t/Zj@ "}BkPMz$@ NhM&Vuٰ.u Am0#!hDžAt?0 Oǩ bY~}tttllQDc0-ـZT*6mxҥK[-/>L&ʂ}8ggg3&--mԩpӧOufڲeK]]hi4 6ddd <?Pd2|-^غ8!=zҥ]͞oZ0K,r @W!l5T*?`F@}}}vvvVTT<{믿NNNk.]tI&m޼yȑ7nI8ݻwFqذayyy;vcǎD"ĉ?Ss26mOJX,@=G򉲲>lÆ ťG;w=Ię3g a,gΜIAf:q //Z9soׯ_ZjРAIII}Sƍӧï]vȑ,Zhر"6W^D7߿РACa0@f͚{o޼yA}G}m۶:@ddd@@#""NJ^رchh*l&6W^yjo`0,={wL6b>?k4m۶M:…,[l۷x<aqqq{Ν_x_Z-$mQ8 NeZ{=FӥKWVrd-T* 5@-?!޹~y3LHZ~SNeffB .\n_x?… f~2eʃjjj,X@ܞް~I*3}twʕ]X,FjRUU'tV\_N1 *jjuHH B>|x޼yBP(=ĉqqqs Ϟ=[XWPv~s;u1FQ.j4p'$I p"##?~񂃃-kdk}=L&BZPSSCF"(%`_ ؙV2o=Hc+?^zY8JP*A<7߬>uTcccxx7x<F} 7o^bb"B)**[nϞ=+W,--ݼyA`ZMhpw' +++F#JmR> MU]] crD&mTɓm֯_Ȑxh?\5Ilrt9tP?T*#i !277 Z!-"4 r\kP=f{J*Ơ^Ferƍ/H8zjTTaW^͍OMM]reϞ=1 YTTԡCӧ'''k4xfϚ5~!P(FVaaaD`ַo_'ŶT*'|2vX 2f̘'N0z eddk0 #"//sMbkaǏ?y_~9m4 `Z7Қ δ}<]O?tܸq/^oVllX,RBX,nUѬSUڵo>}СC777@NSSSN:}t1|p\X^^>o|cǎeeeF̙3A={633SVWVVZ ><11ƍ ㏭y\`ȑ#ѣX9vfpݺu[SB___~E"0 77(!܀ vjjj44{G!ls౗ו+W`$9}t^^V<<<`úXXXرcGx~ƌnnn=z`2999ׯ_ k3V^]TThF |ƍS*Z͜9s޽p&...//P*;6220iҤAad2]]])SdddLwy'886N:]rڵk W^L&Tӧ9agϞWnnnSLS9s&11>l9TjkDnl[FJH&Q f4_5QQQ֟:t3&ჂaNT9&ƃBئheС3Lp'۷mgۻfYVqH$(),D;v~DK,AM+fjkjj.ˍB w*h"{kAlomG " RWT= k-%DoozZuʳ aܹslVjZE"*+LfJ!QRBQ@dgg!h4Z688TV>1l BGc !-  .['C2kp!D 2 qWWW3SI?%ku }UjCWN'HF5UU* ٫A<;YiF)//71A!D8'y>zyy N0ssǞAAAK.͛M(_|s,X@" bƌC޽;.q˖-AAAaaa]t1bmOݻw ޫW&J8nk ڱcG{̘1/P۷oȑ#7oܾ}KfBL8Ѻv9@خPnQB@v11Vl6+={@Fnd,qC1 fX۵Lj`%X,uݩh*++atرNפW_M|xCCBӧl&"..\8|0t`v^zM:uذaG3fǏOMM;'OӧϬY222;vsfff'|ܹSPPfeΜ90a$8pÆ +W$b޽111SN!!!ǏWyy՜h4Nhc*:}t|?--WZ <x~~[x0;o"ٳΝe03g\rի9NAA#.\X`3g Ä vڕDV%b۶mp?|ӦM~֭[׬Y3p@ Ne˖hfÍ2w#sNXXl|vP(d2٭[?{k=f[h06h/^HPD"۷0 h4T*522r^^^s΅q=z'ի}˗l>CjHt9AX,Vl6>{ni i ٚrǎׯ[4-8FGԐ-Wo44 yCd -OKKHxn0X/9s.]:c x>77wѣ{= RRR~&L ENNΊ+JJJlvbb"fCS H.]`=..ڵk)))MJ&-[hllu4l0]0a Nd2_۾m`UKh j*$P. FxNNӱt_K{뭷lْlwFLMM]t)=zw^S[[ۤܿO>9~xppÇ9B 믿Κ5k{* RtqqyAnݺޥKgsz+Vi&N6euĉ>}ׯ_:ujM4 ON6 M`0FWW϶ sj`&>ΣNMV(M#R`0 ٜXtE֬YĉZ'%%aիW'&&bi0.^Hbh4AAA0NX6cƌٳg`p~V:u(0 5 *ojy:ZҥKE0 0aBAAw}g2p>x aT*`0 2Lm!`6dsժU)))111999!!!pR)Ld2dXRRRn޼߭[7>="x^rWƻKH[֔śӷtZR2 ___ S60KVi8ۆD6Q>--!BǣGԩ@ bRPHSTJ/^ypq-!U'M #f 뉄c2AM c/ٰajgl6߹s[nVMas8NeBpܸqaZ!1BAeyP&k9 ҧ&z=P(0D"еk+hgOTSN* g5xI[%h2F=iZ 4[B!NVP Scbb.]{7|ׯKT*m]@kԠdXI2 ۷!t`pWϮPlTUUݾ}oT-Bp?l0d}Y+QI$zYIjkkqww꟮## H V\io-Æ sD+ }GKHRQQa aw}l}ivESJl6b.WXhmP*ZT_t!c]1Zm]])!mA$/hFϚ5kÆ RC -d˖~8rbfBFq>jZR[%8 YWG. X, 騲KJNϥvפmߒN5:a„Doo9s?ugZzVh4\.N-F£t(h4n}uuSjF\^<:%p `\^-{$Q"}̀SM 6AFFFK'wԩa&A! (ŗ ZW7Kʢ V[Katr@I dfm AVt\X"Thdvip1BMt>hZ-\D7jJ6߇"555\.DK$b!$kʃ}"Jn}-v['n]+;{_m>M$-hҩffBX,Vjz7 12U/ DAN3,UPUK  f^7j_۝rWjB4)wS"V.]'[?j7!,&HO7`"S5/'\[]o#Nb $m-r G40B`M dΉ\.o$lWӫ4:-B`+%Q *3* ^ERDA( $HcGv CEcŶWYiyD<}zyyh4ZLl//ExdpROT&(B3cO>C,vyt :a.,w^e)^^]g=фnݺa֢5a\aӹ yyy08JLtrC&Q8 ^w#ȳ[7T32jo{m۶?vssSt*FYtBkziԩ :Nrooz'5?޵ka): 5n6@oڴu]7n(7~{ZY(28+u"gT*0`d/zѯ*hNeްaFޯOW:qɓ'NS(#G,**:wAذT>.l%ZJA`-be~A ,8rȚ5k222Hee~wbDNAtܑA:ww;۹+ˊ.Nr^E@=.U<rh@C[j Cӧ&O&MdwXJmгE-Bswߍ76/--DDDQ]m=[TFr:9`hѢ'N\_ҹ\˦s"ZiDh49rd̘1ѠБu7o$ml)aRe@-BsrcǶܣ/--R!!! 5xHi) IaXVHHH߿Kˋ"oR¼Q--1z+%-Ia5. XUC\Re@;#鴲垾l.++ J?ِt*`T:9j=a؄ n&:eK ɶ&-B@m-AR  jhSCڂ&//Ivנ٥^s观俗K /lJeQ6ֵFeLz  .jZZ=vXtE"@rC/%a uҪA~~nݺq- Au5o %-h$ 4qFy[ ֔k֬!-X,V*5ȍ=5ПlH:i0։0֫wwuD `۽ܐKb/_:iǵP@ڵ+0kld>v/mA%Ivנy\.6]+$LZg*..oJ:a0Aլ,ٜG*5j-6" CCC@ gjk ڵkW ..*Ψ5Y mAgmAN5kT,oٲe޽LyD!$ϭct*d1(2A2yWf̘q}>ڿۛ'i{p}Z)6O]]Ap(Dm-!Poݺ8z6BWjd<_{{:E}l6dAAٳg+DnFJ;( ,WK&hr+11֭[(RьLsεBB)ARp~T77J34w%gϞ-(( =qhLHHPTA(J8:::99?n z":ɓD #(( ~熤S$WK .ˬ,s]]ѯ^!{!h4/j:#h^$ úvT*=zaPjT6c&''GGGtz`04&Ν;>|ZNMM;;S(*v"anfsyyyXXLTT熤SY™2φ2d<~(딂z_WJuCrIAt PXXHPaQT jӌ1ZAD<,k?ۿ#i:BOѣGFNqPScD2̸Y@1φ uנqK[(=$_ [^=+m[b+&cf;q]O3JJ[/qi 4-55ַGU`cuV_qznH:i2u}wƆaXffKN&yRUKH6gBZ-BM` K. ??记TڽdM^^:dgꫯ r!k >i2g)))̔6X7W6Ph^M{ȑ$: 2{YWjk g2FN@-h*lo0Yz}~~@;ݻaXxx8NYF ԛP(ٳgM͍lX,p8pUtHk׮F~ Z7j???1+![d|쵮.s6zyh縕Ne1^ȑ# BS[[jT ` bn9i4e2p8 }s B!P &ިm i tYZVJ^wAڂHܽ{7\>Ѽ1j>FDUOH0fB9p@$5۽|?CKVU*BQBiibr)U:e[(H[c ѱc&'-u^(ۏfW^p6V;!t*FaОYMoKxxxPPPumu N5 ^3p4T*ZeE}WCCC9tD"?b70>{F{ aqڌf8SF҆-?Net:=99yǎ5 &BT*5LЭLM-Neh\AÇ>>< tJz@di?٘7(((yD4# ,>ьXN`RrED4 R;>sX~R$мj#&9{~.",t&J&cbb~駪;xԃ:w/)#[T^t >}<{><<|Ν/s{ (h/fph}zӶNe2hT#55B!UAƶeH$R@@@K?#Q DbJ&ڵkv Pb28"\dɳ'y<(yD}JKK1 0L$\.,TX77=#srz TcMMzغu1cBRU1thKN%(WE4hsϿGD3RQQі BCDv_P}"992//jiphTlʈDhI۷#;f͚  %zā7'tcz/*0 sj`&]b4op( Ĭf<-- DKިW%NDKt* 0oD"mܸQ ,_fm-a6>BϐK'sW{A Z~gUQl󮣢B$p -ᾩKMM Ljl$Zs&' 7tիg H.#T])=sy;^\[Q~d֭[Ν;DĶmG BaXZRH+f#RSSMPb@j<ܞuȑoFUՍ;wJd2ieCٜcFa{uoR[9 c# Gxٳg[#D[d„ ;4 tq#_,aaбWCBMzz? 忦u}#G4t:}ر-bVU41 [v+W?>kw>lhL%;͎_[ h3YZhB_Y{;C 7oBezѼ`01낂dAp+/-aʟ.~eQی# ҽ{MB"f`s.믿ưon&ݻwIIO?Hš*Qsp CX]]=eʔJ\. ܹs\\D Z*۬YP Bmh޹y>>>117nԺ<͇1ra2 9RΟB JKN޳v[2+WtORh^O8dh,8aN;lذk׮z>---33sԩV F;fk@;?{vZj$}Z,ǎf KQѤo E&a5f3QYeV1h/ð,Tʺudee9r6L$&R*8!$$$>|`꫙ϟxȑw&h,]biΙr999xS]붜P4/drڵk;xu1 %U?ߣV\٢IXNT@`_҃ĉ$l݊/ZL^|l^R]NoD;?tׯwoٷoٳ%B]l\OTLJ" N^ {>xŇC("&SsNzkP"*Ϗ L0BaFZG !1 j8!X,YYYvھ}Y6m4yiӦmٲު! vâsZ(@гgg+2J f:"Z_wOŖ,!Z2)a֑BL9Jemc:88xǎ;wtww GaV F2dH3q r0(YX,lf`[bq{uxo-@Rcsq}&)Un [Pߋvgg33;3pWWWkkk(ԩSv{xx+,,Lէ=m+ϟ4S r H{qm̭HgTU"z ¼<}}}==25f~++B G&7 Sj}>*B"xƍÇ9իvرcGTTgϞ]XX}!0 :Cda N_$=!LE555555 XBgֶE(p¼+iS j*Mn#^">}};SN\.[:TI$۪K!==N7#SIzb1`hg7J۵k_] ΫRK_/|H" Fa|vl˖$OؑDqz~;MEB/#Fz(-]hĉwahh pww߿^s/ΦSEE $䲲7"|<ɸ~0ɩQZ_zׯ_..^dfH"Lum|ĤmT;Bd*ڑJ w/)|!@] mmHݻw[h7 To6㙙VVVt:WcV3"0ld۵KVSQhmX\zm:XElٲQFyzz:::8PюTRLB }nPZ.!jfH-1h ,Ԏ`v\,FZt=U.oP!J655m󏲪}ŋgnX[ ovj)/_ $./xuC8u,^7sR"dA`kVΜ9SKK̙-XA\&VEmPcj۶m_>}!`0!Ih48ofgMv%52<884O8 76I ٳ+E!tvv.-EDc]öYCCc,Ν Jx"yEiWU*uL~ϛfYZZ"G 5jk*i݇S$Wf:]qUxYY׈!ebbRR^eӚ9sڑ#Gtu%sb~ UJ*}bwL&C,ӳ#f608s&96l7sID_RC̉6ѱjyҗᔖ0̢m+B;;ƺ@@]]O;U꘲YYcTU^^^ՈJg}ˑ2KC]k=?<<]6Y}Fի8_\YZZ~+**X,3aBF&] .\!tA heR1 =6MNUZ|gO(/G HNF oo@܉@X?k$}hmakccUQZn+JMMmѢE_#LTKvm6RY...EnnXJ neXRw*BTPF xBBJuq\Sj}%Ktf zX)#1o_LJ)|&:Ŋҗ(33RQD+0 [`L&;rH.;rXz"TI'Go_&)sssH<9t"*[/Scǎ,_r)..>n>C'IqGSwU9H5=*==йsLD֙ehƌc``p9]]TCWmM8W{"Tcɓ7}p#8x<`0X}o<]*]믙^K׷c}McLP6P([ɓ'٥v鹹AG?H۷oɐYS]^_hV2,% aU-ϡk6à" \]й2NPU .|&ImDUYL++!tկQi\ B&R`g-VxNu2w:bM-Yv0ҲPUNo޼dNN.׮v)Z4k@ <8555)tYSw1׳L"Vf*B !}拓{./ʜsdHdV] H:2J&R-555*jzI,<`ɓ'%'J>2T 6XWKhhpkGwH$,ڶmܷg<#3T}^6+WT^^NFVll%K2t@6h_MAAAlL *禤lx#Az~Cm6y25~jjBcs{Y١Z!CLXK$\.wΜ9nݺpBd9{S7A/+C^!,/ZiCT] (z½O c`Ѩ88r0ϯ::-! XZebK(bbb0-]='CyϟX[ΝR)0׳ r xyZ>KZby[RH5o-t%:rG[nرݻI| L_ :ҒP,`=+aǣD"sNƙyf ]A }YZ7mZ.y`<uϋS.Qk)W"p|*>ܽ@ԶgLMMݺ5Сyx+!:"hѢɏ;_6Iu,&䖧ꮭ޽{gىbx+T^qZF$B"^ݡpu{C!H?<#)/RÇm[ccٳ%fѽB0-S4ͨӧOhTF،HKJ& !؈lo.Ŗ- ̘qK1ْ%ӥsHGKX6!&p#^paӭ[hҤA}z.ݘ?$V9 t:ײ&Or#ݻwpΏ+*LfkDIFUS)XlF @ɓ'iixD/I w'ڰaCJJ6SgLoH$nBX(/!n@"R^F^r)kaaa7oNH–.LM*޼ysرF9, w6r>'|J.]޾} ,RIm2XG\- Xx54Y_e4 nC "lB54 CC1c,tK bep{SSay,o{wXgN#u Ͼ(~Flr)}^iKz8*0E:RWw,4p>n!Eq5ܲH2QءZ{ݲeKYY9`Xl4 C8p`ee<<aFFw.\uuAZiiiFJgΜ3f tzHHW^YH$"!deeٳ*ȩɤ O>H${Pm"ӫBܹs5T"7 [ھTΟPu*e𣢜%%%-*,, y z޼^n]C-immc:tgSbĴ705Ln\ʐ-gjΟ??**jƍW^q<66`g!N***ϟo 23[񼜗<@fd!lhh?~<޿s[cf3]-!9r$aaani[<0_Rw<`EZMZt}ui 8t; kmLOP{At$J҄;͛MMMk׮z]di`qUN2elo{F׹%ȥ𴴊EH !LtttEа}v B&ϟ?/᝝7H&kQ2:vL@"̝;7++k޼y3f̀=͉ׅNEo;rwSqMߙvƮ6lLv횋  KnE:ǏkhhxyyՁ&aFe ԩ &]U(2o6I'Ovq=~e HDq,BM4 B!=l썻(CB?~|ҤAMM;ʥ'Gjաun6gpW3e&M %SЕo~*o>qq? ؟6n܄t}iR<,88XPTu 8k׮={˷DDD ?"D"fSS<ɓ͛3!A6#P($L&GEE łqN=2\Rޓ[/OoCR.O,gϞca!!9lɁT_I{WiiӚ7oÇ:e޲f1ͽ)*J6y``7ԩS%uIҲ%LfUUUCoHѯ{/JᰭTPwj;E{vDDDDDD|3m"bjӗCv6bPU02RrQZQTRuà9 aT\àw: \Q׮V&H B$KB'NJ%cG?ihΜq"õKpə\LiiiA?>J.;o2""d5oJ %;;[e"۲Hٳ%gp7nqyoquζkol|OLre:jV|IlVIO d^^2uVkkAe. ̽W.LT(L1r2\_F|@ x?dsר=p&˖-ϣo!'O3-EG?86& k%&&7n͚5><ʳC=Logj ƍsrvʮDbs>(B ]] )>}mDGG!9fF4յ!TSyӃ0/etJxmGS8{k tx,, ! weXGeIU=m=]Q׮] zjz:zZ$''7СC%%%pe*vwZ2龍\ LMElat41qmDĥݻ=z}!t] ޮ# v;ursv4Ħ|`"[;\W*n޼944N)"e2m8QTVSʐJ5Zf ZnudDpoxqMީ$oر~TfaJ1I_[KkkO[{İ?8{)S/XȈdN<;eo~k`R_9!mlM2]0:Z0'g.]^`0ͫL-O9mMc*_OYL/.N3}F<xzr |Lʯ8cƌuefrqQ[e4 %鎎ܸϵ9>Ojd$+[$\oasy.ٳܹcooi޺ޔHT %aGc99ikT8y"%%eժU򲶶PSfv["sFO8x= ^Qp&8ʕ/+*{߿аXNE3:9]ǎ}0u4Xxrs=Z)JA!5kVs;}ܹsZZ3Ǎm΄ s=MAapmN Çc5kffW?}*ߘ9B28l_P(SLD+Wp 9oax ̙Y8N8hUW[j(C*>FF_?~]cǴL7oݾOxտezKlm*6$`4`J55/~qذi}###?I-"ԽW ׮%XHքݴYH$ky [EX)K{=QSN8@:tDlȑXX% -+L2kffOә"֋v8Wge,jɱ\.`g'MRPVUU <`5պv}׃Z^_|Uu}6!4҂pϞUTckR ֑(Ξs"AYwX1gEA37LL&6 ?c޷ؙ~xMM^mEQKve*6l]=8Xus'(yU$ZB^^ïjihoLX*쐃m7paXg;K_^f[ajj,bcьz (z5y~A`XC IMM]pAbb_~!Ca^E]̽x`#H$EBMc\R%HЛ7lGFF &lkRt\zR>rhw;e崰nHt>Jzh(}udB{{{ervy"n񅼼/Nn@P|/_ muu TWW8A<8G$Pݦ9vxrE99H.ZH<OݴiE'KnݒK6)gݕ%r:M!Q.>5"C~]k]<]X$ ٶI<9vB1_&PzׯeOʢdW^ 8np-N!4p}.CټWLsHҔG7oSQǓYV?5h-q0԰\\^E6n7G_62T*(*BPߔI&=ydȐ!G9}gYQO3{MbrصkAOn{fc ի8.۷74q١d䀀'뻴(C>.;l;;B#- =,k|;j]Èe&y-q u vmmxz}OeIˆ|>ْ%҃^S!ޥNFF /(1 ާO .<@sDޫи7&\jpߞ=/Hvz[p?׮L& ,11;00PM_3[]ͺ}*r)],,횋06t*M+Jeǭ :5TH* V_Þ=wx)5wEN6,4t uuXYA.>m ;T bc*bA2 rs\Vo#BBBN>Mя{6[ +'.3յk?EG߉ixqעduep̬we˨|~92?R,N\ݵ7R31NZ%g{L{o~: u H"ӦMc6vi]] a>9O (**_B"'Aʵ,;;p)CY9%{7~,^\Щ411 LN}hٲe/^Ѩ+pqM OieevCiwŗJC !C+wƍ6vM7Ǿ Xʥ NZh迅atfai: =0 X,VpM ͍u,+g۔$FQQ@x12y[a~뇩),ږb1ؽ/,DFƼ0hcŭ 5=<;6 q`v|xHprR2,+$J1ܔr?/]JHN7dȐ$46vq9̍}2Uv7j{o4Ե=H&*.I$_mmTȉa͛W^X/q8f%yrsS߾MP ³֕Vq$AiRoC h6mL01ͮL&uV׮]7nH"/_;xpmp>]nՙ%}fh e H8y;fvg?@)ܕѹs792jQ%coܺ:h}(LJaܔ^m\b)28i9;"##t0L|.V~ʷ<ϺhuWw!RRR~qY/qDNNGG9u霘 ^z7X怯yHLLZ[GxګWlaܹsB_%V10']:JWBJ8҃}&脭)-[$dWtaX ޛ'kY`gHI\-*T-B">sT`jbn:F 9ڵ~A4Ttlxׯ<<`tVXcYYB555 .twwp|B655;ǎ-2Dd2T`oo7n=jjj A k낧;Wx*ff}H9#~|UK&l6ڵK24tI\ ::{S7sz*}* R'-g5Ʈzv)FyrTCD"q }Nx">>>##w^XXT*;vlzz7v.e͛`S|Q})Z{U֯_!DuF2 7ήXt*ZVWfUg;{9W話O(t h(a _diRt!q>fKKYeeUNmcccSSScc#p8,K,hck!W,x70:2-߶ҥd˗O~&5!H4o523IF)Iw+}G2`nR/+(4VV**:֭ɓjh)ѣS VNLMiWV#\`Att4jq=J$#F?Ǣ`FuUo8 gbx)uy;))lE(i^@1 7;v={v"" ^[Qm%ct]_DOi4ڒWG߭ǣ'T}BΨ%X[ gdttF4s٘>2ad-Vj8Q~`n ̘ak``gee,aXYYaѣ:IIn܈d05܊*Ce\$*Cm܆7nܢE\(&3oNi]|IjkM!ɡ÷l&@*K*-j-ץ3rҪg,=P,ǰ =u+Rϟkjzfmɔ"H ݻ7o޼;w2M~?~^OCY9GM @N8qٲ2.Oמ.\P''O?4T&E}jm؉IQ#) VqIm>(D-[V^me>Ǣ OvS5^Rd2t7oOXg_7To M0N;v̙D"qf D PYYŋ^z% İ.]ܾn``zmmयi矸&8Bz,o! l$ZZZtuq*mp첳utt+ܕ2P$M06FaÅ =יRx)(9o߾ 1tb>6kppLnjSQQtQ+++U]RNŌ˯x/C9<*Yy{wUru%KbccR)Htrr0aP[JV[M斧v2u:Imɱ܊ӷCs҉HFåF$AWWk;t=Z;ڵ<OKKe\p%O}q @%pC ?*"p\&x6n6MB1_(拥"B8yY1L4nJKR) 7lid Ǯ! x$>xسieLMf||g^|ٳggg*XYsDvիWF_dpB BNEoa>DT@]Ut)t !RTu UUgG$4#du*agL6ELZa۱g)((زeٳg---_ŗq?"h:|gCo6GΜ9S+(>j#N59U{rR ҊsSM_W˨>"@"$-XOph?C]uǒJdž  5k-[`#TkA]ݲ 8B IxBKO~:-#D*B฀^1X8. ނ#YrS.'$th>qĪU444_w7͛P0|rNp%eUl+v++NL\]V`4 p(lh6E׉_GuRt (ٰtqǦ|wtWTBngH$T*y<^IIYQ~'dulh:&g@ B@SSBH$@*]daa99݋N".g/)sg8d>JW>*&M҅ak~lۿwHSRtO qzƊR(492ܴ 艈ά1xǎq/I#7dop4gea...>>>ii?0c2!B:mmn'Zt !d `a՛&\leLT;,$ݝ MFR0d`&! qd1 < fPjQ>xN.N}A$2Δnvuha3Y|Dl0_iJ[#m }{_>jesdaTN,!);i ׯWDMEQE2ӵ2pT X5c A @iRtm!Uc@mOm^ne];T ۷("EF˱ W~ѶF.#ukeg14o']0ձҢn8 ;y>>C[NGס{նC9::bp'*a 5jp[dlF#\ %ǣDBj=Ɍ*sͽda^u6*MzUtb:H$wQUANZEըQ;&>|8tP ~TB:ݫfYlgLMCa|!mĞK,z?N*vɅ'{y| ӉX,~|θ2R|nr)Ug!^SuFpFP,];W1-' O0H;W]]_v3rx4`~\.^O3o'}K Tj+IҸ8EP-U *)J"vR[JUJ*]r)UTcP.epll9C| gGF>ڵkORJ]]}ʕKjwCTT<*#]Mw4K"(V"E "b2:DI&稃ѸQј#c0Dd`@ 4 on^GuL~r֭VgZ܄L&ʹ #D o9oAAA{3b&r'-F[V! s!LNNf)))&&&f5B.\@x/r@H+W>leeс⅃ޱs! bggPJrCL7FRo/RhS剨j kV^=\.P(Dd233k׮?Y3G}4* &i-f#ʹ?2 Z-_@d2׬Yf͚U @ ?xuxjzppАLߥ====dh{ L͵rrrjkkKNN޿!t0RT{ T+rtq|ddL&ϛ7ozӍrohhdNuvv[ZZL&p8?Յ o```b<%:::TtiI.bq\ tvvjiiMڝ$Ȉ ±1r߿/MMM'>@޽;::jffFJewwTV(MMM<266RӕE&wuu*&ꪫE")VT]]]<000 SPvSSӉlkk0lhh`LJEd2||vxb* W8yǎg777[[ۍ7͛7믿r ߿rʄ+V:t_\\pBss;wN} %c||VRAD"Y~֭[###}}}% wwwY>::߶/ C<<ƍ|7n8pڵkX,VttRqcҥqqq|۳g\.[h xAAƍ{=+υ\.x7r9gΜ ݽ{U""" \\\k׮(JU*hhhttի766P(/^L')a&9222 ȈD"$BPA`}I(r===uuuDC( MLL`lddRt2L"d ؘRA hBaCCĆ0--- }```ZZ~2L,CפH:22=!r~~~TP!"sK6bZm6!0ɩEIR(gX8.,XP]]Ν;?aؘuKK mǎEEEІ=z4!v/|||& Fh^OYP/tqΞ۷Looo<< 8. j̙3>B!H$ qb,///$$7tuuNQGGNL3 *m77  foذ_yy9̟?aN-ܐ͉pӃxs><<_{5T*4551LSSSbڤ6a~~~$`\fRiuumϞ0 d2[VV&\[[`0LLL8u5E$544xx ~~~0@^^{֤޽{: ΗvT-,,|!N _P(gϞݼyXvv6䄆>1ȳof`3Lxmll<<>xy*:1g`<~͛7]qqƍ濷W&_Nq[XXLֈͤWBfxwv@WWWXXǃmQQ3000GEDDV𷀵\.xX<::jeeEP>}(33.,P(/aurrqׯƽH$BY`OmmyT5LRNNNK.mnnh4!!!D@ `0kjjD<ͰUu*###7o۷oyIɜloo߾-+**ݻhhhزeKll,/))iiioڴ +))MMMrʊD" ]zL&/\%H$Ǐ KJJ~ǖˆ"ߗ奥REDDdeeUVVVVVfggoݺU"xyy|~PP^iiiUUlBqΝ@AA$kx=266i&ssP ^?#\.Vm*~ԩO>w)|˗/9s&>>dxxD"ƞ={ƍBǧ){Anjj|}FFƳ'***33p8h@FFFuuu[[۵kWZ>|xUmmmAAA{{O>a@DDĿw\~}۶m8[L&ðp8{as8M6^ztt4((nה´~7nnnFFFiii666 %77.+LLLMMM;;;-,,D"QzzׄIffP(o/_z1@jje޽GR\.~:`ҥ, ;00|-$kYd a-j4x=<((jIӣ}].\xJ^SS3qNܹs' fs܉4ޚ:R!ͪbeܹH$G4A^^y \.gY 7aG9re󐒒uV8:/-ZA1 yZBC@ /f"RsLT*>onn9g, ###_O! `F@ڭi9_yfx._|J---'.P(uuuhb1 8;;kiiԴjkkbѣG:::˖-# /Wigb:h+Vwާ~W_޽lŊ C"[L&:33g{{%K.^c#H\.766ѣD;rHEE{ggkkkE"ӧT_~`0ܹsĉǏ.[liiivvv0NJFјLfoo.DO훛,Xp6scccǎD~bG# @<ؔ^oooHki4?III%''ٳt޽{rs}}=aJb566¡Ѳ2wwI{pe25qajZ$ݽ{ðF>~Iqfdd30 ;p@HH$JGGG1 [nƍsrruxxx^^ޯ<㓗ggggnnrl;Nljj B}}}>ݝ eoooyggIx<<<#""97oLII144TPhcc3O [__vZ8*L";>>N-[FJΝ/?nii ~9@F!T֭[w!+++???}||y<q\WW7..6L& B:SԞKK˨/ׯ_g0o?wp8Gv?-aMHH{yKx INb(éS? F]t )p@V SfX$, 3e˖CCC:::\.399D"QZE-LNLL\z֭[R)F)++[jP(R F 8hjs0 ##b022fff H$ h4Q,--B&Y__/ÓT*FkkkxӔL^BRwfaaʕ+WZ kMPPT*J ...W^_vZtҥy222:{Çդn,k޽$)%%>H$~ixx8N_IIIYYY.\Xd Xaaaaaa<gee^722qѣIIIpaɓ'0 9wz8 ~#-_ ?K7qy?ϊZ=M s 2ә_K"J`j ?_ ~0 @V' endstream endobj 379 0 obj << /Type /Annot /Subtype /Link /A 380 0 R /Border [0 0 0] /H /I /Rect [ 380.6723 214.3284 390.3096 223.1467 ] >> endobj 380 0 obj << /Type /Action >> endobj 381 0 obj << /Type /Annot /Subtype /Link /A 382 0 R /Border [0 0 0] /H /I /Rect [ 35.2500 428.7592 485.2500 621.5092 ] >> endobj 382 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/files/2014/10/nowcast.png) >> endobj 383 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 384 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 383 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 42823>> stream xw\Ggw:DQ,HQFb4jby%c1D_[)jX[DEPAPQ(w\z11&;|vggi`A@ +${+@ =A@ dѮA@  ,@  !@ 3E@ v j"]ZhנYhא!$H"d@  E DBDBVWWfSR) hrARi%gϞ@Dɓ'NܲeˤI۶mYvm+$-'Lpqȑ#={tuum=M|>`۶m&M RcO ʕGf2fl6ZV FTTJUZbTTTSTLF"*//DBC$t:K. `6 _HHHuu͛7drppի|>AwƒdduLhhh Ph4l.--!!!~~~FQ?~8!!D"ݿ_*FDD[(++^^^aaa4MVL`R͛7T*Upp0D.ƲX,@eeWqq{HHN+**tH$z!,&l h4\.:vhU,+44>px䉏ϭ[\nTTUrөS'?BPє(.Ι3zjС]v&);;tѢE.\(..NOOw333322ϟڵnذ`L2%,,F*8qfO6!,,l'NP*Vp8_}Ո#QTTԻwu֙Lcj Ry,̟?q*zҥ;wSӍ9D">}۶m ^xqAAAtt}>}[.;;W^"h}IIIܹH$:x5/d[u6o)ziii!!!^^^ hF X,"رc<oΝ[lӧύ7Fm[ "---::Hҽ{j41c0LCOGK,ϟ8qbjj^ _o&%%ݻw/<<믿q{ɞ999o;s秦T.]̞=ū=#m~aȑ?H$+W3go%Kp7C ٽ{d _n]FFƼy,KǎE"ŋ Cu:R|W9( rj]]]@@gLL̽{p>v B (w0؆ zlFL&Rǎ5kƍgX={B%oݺաCJ84~O>W\/,,ѣdzUTT㬬O>qU\\x@@aI&m޼Ϝ9s֭2 񆆆 Fc}^oF,cvʹ4 lCQT Êl6aIII=RSN?~|VVf3l͞=͛-7ۘ|>_.h4Z\\z~aСC Ÿq:~x\^]]m-22o߾ ca {)//>|O>ҥKaGyѤe˖¾SLfϞ=alnnn!!!766QFm߾}+VʧcH"v$۷oVVO?%$$>}a>>>_~=B5jԾ}l:t0k֬4ww1cƔhѢ7xh4}wBКkqJ[nׯ0̙3> 44/T( .lhhR -իW_~˖-틉w$f"Ç7|>L&}ѦM(JBBBRR5~ZmM?СC4ĉu:… ׬Y3fg?~FWW۷VZ{ۗݻ6 T*_~$dܹfom 5I'bM,wڵ޺ Xq={g}fo?9ٳgȐ!>>>VᴬY&??`0teN@Eh4uyݻN@+`G&a0C+.ӧ[sss-Kbb"aR4==}ذaV ڜ:ujС᜜44رccƌh۷ogggwd25Qٳ*jΝ...O?{ȑx߂t:]UUӧ?~ g՚ܻwoذaݻwoղCj"|Q^^^w_ pD|&IĢEf̘N>=~ \޿R 7o90 .4Ldڴiׯ ۽{_~ uVfff@@@}}~D"FRw}w̙\.~Ɂo~MJh4*׮]UUU>>>hI@-BsioMyW6/ ///p8/S[[O< ;w.y}-[^#֭[G8qs#0l˖-t:O?]hѷ~;s̉'Ѥ_rǏ bYr%t{&;u$  4h+W\re̘1/_>BJdffԼ|xzVԥK3sѣGCChm)VVVFFF0 xɟEZ...XYY 11 ?d "2qr7{7nd:tСʨ( ***_ϯ{]]]m$;V9jUk愻::"K~'hG&LHKK5kP(Dzxb:ѣk׮__8>L[&O>>Ph/ 9 8rn̜9s9„h!~Z Zy-[wW.^F8'h@ "z^:d1%i@hf@ / 2D3#4t/Zj@ "}BkPMz$@ NhM&Vuٰ.u Am0#!hDžAt?0 Oǩ bY~}tttllQDc0-ـZT*6mxҥK[-/>L&ʂ}8ggg3&--mԩpӧOufڲeK]]hi4 6ddd <?Pd2|-^غ8!=zҥ]͞oZ0K,r @W!l5T*?`F@}}}vvvVTT<{믿NNNk.]tI&m޼yȑ7nI8ݻwFqذayyy;vcǎD"ĉ?Ss26mOJX,@=G򉲲>lÆ ťG;w=Ię3g a,gΜIAf:q //Z9soׯ_ZjРAIII}Sƍӧï]vȑ,Zhر"6W^D7߿РACa0@f͚{o޼yA}G}m۶:@ddd@@#""NJ^رchh*l&6W^yjo`0,={wL6b>?k4m۶M:…,[l۷x<aqqq{Ν_x_Z-$mQ8 NeZ{=FӥKWVrd-T* 5@-?!޹~y3LHZ~SNeffB .\n_x?… f~2eʃjjj,X@ܞް~I*3}twʕ]X,FjRUU'tV\_N1 *jjuHH B>|x޼yBP(=ĉqqqs Ϟ=[XWPv~s;u1FQ.j4p'$I p"##?~񂃃-kdk}=L&BZPSSCF"(%`_ ؙV2o=Hc+?^zY8JP*A<7߬>uTcccxx7x<F} 7o^bb"B)**[nϞ=+W,--ݼyA`ZMhpw' +++F#JmR> MU]] crD&mTɓm֯_Ȑxh?\5Ilrt9tP?T*#i !277 Z!-"4 r\kP=f{J*Ơ^Ferƍ/H8zjTTaW^͍OMM]reϞ=1 YTTԡCӧ'''k4xfϚ5~!P(FVaaaD`ַo_'ŶT*'|2vX 2f̘'N0z eddk0 #"//sMbkaǏ?y_~9m4 `Z7Қ δ}<]O?tܸq/^oVllX,RBX,nUѬSUڵo>}СC777@NSSSN:}t1|p\X^^>o|cǎeeeF̙3A={633SVWVVZ ><11ƍ ㏭y\`ȑ#ѣX9vfpݺu[SB___~E"0 77(!܀ vjjj44{G!ls౗ו+W`$9}t^^V<<<`úXXXرcGx~ƌnnn=z`2999ׯ_ k3V^]TThF |ƍS*Z͜9s޽p&...//P*;6220iҤAad2]]])SdddLwy'886N:]rڵk W^L&Tӧ9agϞWnnnSLS9s&11>l9TjkDnl[FJH&Q f4_5QQQ֟:t3&ჂaNT9&ƃBئheС3Lp'۷mgۻfYVqH$(),D;v~DK,AM+fjkjj.ˍB w*h"{kAlomG " RWT= k-%DoozZuʳ aܹslVjZE"*+LfJ!QRBQ@dgg!h4Z688TV>1l BGc !-  .['C2kp!D 2 qWWW3SI?%ku }UjCWN'HF5UU* ٫A<;YiF)//71A!D8'y>zyy N0ssǞAAAK.͛M(_|s,X@" bƌC޽;.q˖-AAAaaa]t1bmOݻw ޫW&J8nk ڱcG{̘1/P۷oȑ#7oܾ}KfBL8Ѻv9@خPnQB@v11Vl6+={@Fnd,qC1 fX۵Lj`%X,uݩh*++atرNפW_M|xCCBӧl&"..\8|0t`v^zM:uذaG3fǏOMM;'OӧϬY222;vsfff'|ܹSPPfeΜ90a$8pÆ +W$b޽111SN!!!ǏWyy՜h4Nhc*:}t|?--WZ <x~~[x0;o"ٳΝe03g\rի9NAA#.\X`3g Ä vڕDV%b۶mp?|ӦM~֭[׬Y3p@ Ne˖hfÍ2w#sNXXl|vP(d2٭[?{k=f[h06h/^HPD"۷0 h4T*522r^^^s΅q=z'ի}˗l>CjHt9AX,Vl6>{ni i ٚrǎׯ[4-8FGԐ-Wo44 yCd -OKKHxn0X/9s.]:c x>77wѣ{= RRR~&L ENNΊ+JJJlvbb"fCS H.]`=..ڵk)))MJ&-[hllu4l0]0a Nd2_۾m`UKh j*$P. FxNNӱt_K{뭷lْlwFLMM]t)=zw^S[[ۤܿO>9~xppÇ9B 믿Κ5k{* RtqqyAnݺޥKgsz+Vi&N6euĉ>}ׯ_:ujM4 ON6 M`0FWW϶ sj`&>ΣNMV(M#R`0 ٜXtE֬YĉZ'%%aիW'&&bi0.^Hbh4AAA0NX6cƌٳg`p~V:u(0 5 *ojy:ZҥKE0 0aBAAw}g2p>x aT*`0 2Lm!`6dsժU)))111999!!!pR)Ld2dXRRRn޼߭[7>="x^rWƻKH[֔śӷtZR2 ___ S60KVi8ۆD6Q>--!BǣGԩ@ bRPHSTJ/^ypq-!U'M #f 뉄c2AM c/ٰajgl6߹s[nVMas8NeBpܸqaZ!1BAeyP&k9 ҧ&z=P(0D"еk+hgOTSN* g5xI[%h2F=iZ 4[B!NVP Scbb.]{7|ׯKT*m]@kԠdXI2 ۷!t`pWϮPlTUUݾ}oT-Bp?l0d}Y+QI$zYIjkkqww꟮## H V\io-Æ sD+ }GKHRQQa aw}l}ivESJl6b.WXhmP*ZT_t!c]1Zm]])!mA$/hFϚ5kÆ RC -d˖~8rbfBFq>jZR[%8 YWG. X, 騲KJNϥvפmߒN5:a„Doo9s?ugZzVh4\.N-F£t(h4n}uuSjF\^<:%p `\^-{$Q"}̀SM 6AFFFK'wԩa&A! (ŗ ZW7Kʢ V[Katr@I dfm AVt\X"Thdvip1BMt>hZ-\D7jJ6߇"555\.DK$b!$kʃ}"Jn}-v['n]+;{_m>M$-hҩffBX,Vjz7 12U/ DAN3,UPUK  f^7j_۝rWjB4)wS"V.]'[?j7!,&HO7`"S5/'\[]o#Nb $m-r G40B`M dΉ\.o$lWӫ4:-B`+%Q *3* ^ERDA( $HcGv CEcŶWYiyD<}zyyh4ZLl//ExdpROT&(B3cO>C,vyt :a.,w^e)^^]g=фnݺa֢5a\aӹ yyy08JLtrC&Q8 ^w#ȳ[7T32jo{m۶?vssSt*FYtBkziԩ :Nrooz'5?޵ka): 5n6@oڴu]7n(7~{ZY(28+u"gT*0`d/zѯ*hNeްaFޯOW:qɓ'NS(#G,**:wAذT>.l%ZJA`-be~A ,8rȚ5k222Hee~wbDNAtܑA:ww;۹+ˊ.Nr^E@=.U<rh@C[j Cӧ&O&MdwXJmгE-Bswߍ76/--DDDQ]m=[TFr:9`hѢ'N\_ҹ\˦s"ZiDh49rd̘1ѠБu7o$ml)aRe@-BsrcǶܣ/--R!!! 5xHi) IaXVHHH߿Kˋ"oR¼Q--1z+%-Ia5. XUC\Re@;#鴲垾l.++ J?ِt*`T:9j=a؄ n&:eK ɶ&-B@m-AR  jhSCڂ&//Ivנ٥^s观俗K /lJeQ6ֵFeLz  .jZZ=vXtE"@rC/%a uҪA~~nݺq- Au5o %-h$ 4qFy[ ֔k֬!-X,V*5ȍ=5ПlH:i0։0֫wwuD `۽ܐKb/_:iǵP@ڵ+0kld>v/mA%Ivנy\.6]+$LZg*..oJ:a0Aլ,ٜG*5j-6" CCC@ gjk ڵkW ..*Ψ5Y mAgmAN5kT,oٲe޽LyD!$ϭct*d1(2A2yWf̘q}>ڿۛ'i{p}Z)6O]]Ap(Dm-!Poݺ8z6BWjd<_{{:E}l6dAAٳg+DnFJ;( ,WK&hr+11֭[(RьLsεBB)ARp~T77J34w%gϞ-(( =qhLHHPTA(J8:::99?n z":ɓD #(( ~熤S$WK .ˬ,s]]ѯ^!{!h4/j:#h^$ úvT*=zaPjT6c&''GGGtz`04&Ν;>|ZNMM;;S(*v"anfsyyyXXLTT熤SY™2φ2d<~(딂z_WJuCrIAt PXXHPaQT jӌ1ZAD<,k?ۿ#i:BOѣGFNqPScD2̸Y@1φ uנqK[(=$_ [^=+m[b+&cf;q]O3JJ[/qi 4-55ַGU`cuV_qznH:i2u}wƆaXffKN&yRUKH6gBZ-BM` K. ??记TڽdM^^:dgꫯ r!k >i2g)))̔6X7W6Ph^M{ȑ$: 2{YWjk g2FN@-h*lo0Yz}~~@;ݻaXxx8NYF ԛP(ٳgM͍lX,p8pUtHk׮F~ Z7j???1+![d|쵮.s6zyh縕Ne1^ȑ# BS[[jT ` bn9i4e2p8 }s B!P &ިm i tYZVJ^wAڂHܽ{7\>Ѽ1j>FDUOH0fB9p@$5۽|?CKVU*BQBiibr)U:e[(H[c ѱc&'-u^(ۏfW^p6V;!t*FaОYMoKxxxPPPumu N5 ^3p4T*ZeE}WCCC9tD"?b70>{F{ aqڌf8SF҆-?Net:=99yǎ5 &BT*5LЭLM-Neh\AÇ>>< tJz@di?٘7(((yD4# ,>ьXN`RrED4 R;>sX~R$мj#&9{~.",t&J&cbb~駪;xԃ:w/)#[T^t >}<{><<|Ν/s{ (h/fph}zӶNe2hT#55B!UAƶeH$R@@@K?#Q DbJ&ڵkv Pb28"\dɳ'y<(yD}JKK1 0L$\.,TX77=#srz TcMMzغu1cBRU1thKN%(WE4hsϿGD3RQQі BCDv_P}"992//jiphTlʈDhI۷#;f͚  %zā7'tcz/*0 sj`&]b4op( Ĭf<-- DKިW%NDKt* 0oD"mܸQ ,_fm-a6>BϐK'sW{A Z~gUQl󮣢B$p -ᾩKMM Ljl$Zs&' 7tիg H.#T])=sy;^\[Q~d֭[Ν;DĶmG BaXZRH+f#RSSMPb@j<ܞuȑoFUՍ;wJd2ieCٜcFa{uoR[9 c# Gxٳg[#D[d„ ;4 tq#_,aaбWCBMzz? 忦u}#G4t:}ر-bVU41 [v+W?>kw>lhL%;͎_[ h3YZhB_Y{;C 7oBezѼ`01낂dAp+/-aʟ.~eQی# ҽ{MB"f`s.믿ưon&ݻwIIO?Hš*Qsp CX]]=eʔJ\. ܹs\\D Z*۬YP Bmh޹y>>>117nԺ<͇1ra2 9RΟB JKN޳v[2+WtORh^O8dh,8aN;lذk׮z>---33sԩV F;fk@;?{vZj$}Z,ǎf KQѤo E&a5f3QYeV1h/ð,Tʺudee9r6L$&R*8!$$$>|`꫙ϟxȑw&h,]biΙr999xS]붜P4/drڵk;xu1 %U?ߣV\٢IXNT@`_҃ĉ$l݊/ZL^|l^R]NoD;?tׯwoٷoٳ%B]l\OTLJ" N^ {>xŇC("&SsNzkP"*Ϗ L0BaFZG !1 j8!X,YYYvھ}Y6m4yiӦmٲު! vâsZ(@гgg+2J f:"Z_wOŖ,!Z2)a֑BL9Jemc:88xǎ;wtww GaV F2dH3q r0(YX,lf`[bq{uxo-@Rcsq}&)Un [Pߋvgg33;3pWWWkkk(ԩSv{xx+,,Lէ=m+ϟ4S r H{qm̭HgTU"z ¼<}}}==25f~++B G&7 Sj}>*B"xƍÇ9իvرcGTTgϞ]XX}!0 :Cda N_$=!LE555555 XBgֶE(p¼+iS j*Mn#^">}};SN\.[:TI$۪K!==N7#SIzb1`hg7J۵k_] ΫRK_/|H" Fa|vl˖$OؑDqz~;MEB/#Fz(-]hĉwahh pww߿^s/ΦSEE $䲲7"|<ɸ~0ɩQZ_zׯ_..^dfH"Lum|ĤmT;Bd*ڑJ w/)|!@] mmHݻw[h7 To6㙙VVVt:WcV3"0ld۵KVSQhmX\zm:XElٲQFyzz:::8PюTRLB }nPZ.!jfH-1h ,Ԏ`v\,FZt=U.oP!J655m󏲪}ŋgnX[ ovj)/_ $./xuC8u,^7sR"dA`kVΜ9SKK̙-XA\&VEmPcj۶m_>}!`0!Ih48ofgMv%52<884O8 76I ٳ+E!tvv.-EDc]öYCCc,Ν Jx"yEiWU*uL~ϛfYZZ"G 5jk*i݇S$Wf:]qUxYY׈!ebbRR^eӚ9sڑ#Gtu%sb~ UJ*}bwL&C,ӳ#f608s&96l7sID_RC̉6ѱjyҗᔖ0̢m+B;;ƺ@@]]O;U꘲YYcTU^^^ՈJg}ˑ2KC]k=?<<]6Y}Fի8_\YZZ~+**X,3aBF&] .\!tA heR1 =6MNUZ|gO(/G HNF oo@܉@X?k$}hmakccUQZn+JMMmѢE_#LTKvm6RY...EnnXJ neXRw*BTPF xBBJuq\Sj}%Ktf zX)#1o_LJ)|&:Ŋҗ(33RQD+0 [`L&;rH.;rXz"TI'Go_&)sssH<9t"*[/Scǎ,_r)..>n>C'IqGSwU9H5=*==йsLD֙ehƌc``p9]]TCWmM8W{"Tcɓ7}p#8x<`0X}o<]*]믙^K׷c}McLP6P([ɓ'٥v鹹AG?H۷oɐYS]^_hV2,% aU-ϡk6à" \]й2NPU .|&ImDUYL++!tկQi\ B&R`g-VxNu2w:bM-Yv0ҲPUNo޼dNN.׮v)Z4k@ <8555)tYSw1׳L"Vf*B !}拓{./ʜsdHdV] H:2J&R-555*jzI,<`ɓ'%'J>2T 6XWKhhpkGwH$,ڶmܷg<#3T}^6+WT^^NFVll%K2t@6h_MAAAlL *禤lx#Az~Cm6y25~jjBcs{Y١Z!CLXK$\.wΜ9nݺpBd9{S7A/+C^!,/ZiCT] (z½O c`Ѩ88r0ϯ::-! XZebK(bbb0-]='CyϟX[ΝR)0׳ r xyZ>KZby[RH5o-t%:rG[nرݻI| L_ :ҒP,`=+aǣD"sNƙyf ]A }YZ7mZ.y`<uϋS.Qk)W"p|*>ܽ@ԶgLMMݺ5Сyx+!:"hѢɏ;_6Iu,&䖧ꮭ޽{gىbx+T^qZF$B"^ݡpu{C!H?<#)/RÇm[ccٳ%fѽB0-S4ͨӧOhTF،HKJ& !؈lo.Ŗ- ̘qK1ْ%ӥsHGKX6!&p#^paӭ[hҤA}z.ݘ?$V9 t:ײ&Or#ݻwpΏ+*LfkDIFUS)XlF @ɓ'iixD/I w'ڰaCJJ6SgLoH$nBX(/!n@"R^F^r)kaaa7oNH–.LM*޼ysرF9, w6r>'|J.]޾} ,RIm2XG\- Xx54Y_e4 nC "lB54 CC1c,tK bep{SSay,o{wXgN#u Ͼ(~Flr)}^iKz8*0E:RWw,4p>n!Eq5ܲH2QءZ{ݲeKYY9`Xl4 C8p`ee<<aFFw.\uuAZiiiFJgΜ3f tzHHW^YH$"!deeٳ*ȩɤ O>H${Pm"ӫBܹs5T"7 [ھTΟPu*e𣢜%%%-*,, y z޼^n]C-immc:tgSbĴ705Ln\ʐ-gjΟ??**jƍW^q<66`g!N***ϟo 23[񼜗<@fd!lhh?~<޿s[cf3]-!9r$aaani[<0_Rw<`EZMZt}ui 8t; kmLOP{At$J҄;͛MMMk׮z]di`qUN2elo{F׹%ȥ𴴊EH !LtttEа}v B&ϟ?/᝝7H&kQ2:vL@"̝;7++k޼y3f̀=͉ׅNEo;rwSqMߙvƮ6lLv횋  KnE:ǏkhhxyyՁ&aFe ԩ &]U(2o6I'Ovq=~e HDq,BM4 B!=l썻(CB?~|ҤAMM;ʥ'Gjաun6gpW3e&M %SЕo~*o>qq? ؟6n܄t}iR<,88XPTu 8k׮={˷DDD ?"D"fSS<ɓ͛3!A6#P($L&GEE łqN=2\Rޓ[/OoCR.O,gϞca!!9lɁT_I{WiiӚ7oÇ:e޲f1ͽ)*J6y``7ԩS%uIҲ%LfUUUCoHѯ{/JᰭTPwj;E{vDDDDDD|3m"bjӗCv6bPU02RrQZQTRuà9 aT\àw: \Q׮V&H B$KB'NJ%cG?ihΜq"õKpə\LiiiA?>J.;o2""d5oJ %;;[e"۲Hٳ%gp7nqyoquζkol|OLre:jV|IlVIO d^^2uVkkAe. ̽W.LT(L1r2\_F|@ x?dsר=p&˖-ϣo!'O3-EG?86& k%&&7n͚5><ʳC=Logj ƍsrvʮDbs>(B ]] )>}mDGG!9fF4յ!TSyӃ0/etJxmGS8{k tx,, ! weXGeIU=m=]Q׮] zjz:zZ$''7СC%%%pe*vwZ2龍\ LMElat41qmDĥݻ=z}!t] ޮ# v;ursv4Ħ|`"[;\W*n޼944N)"e2m8QTVSʐJ5Zf ZnudDpoxqMީ$oر~TfaJ1I_[KkkO[{İ?8{)S/XȈdN<;eo~k`R_9!mlM2]0:Z0'g.]^`0ͫL-O9mMc*_OYL/.N3}F<xzr |Lʯ8cƌuefrqQ[e4 %鎎ܸϵ9>Ojd$+[$\oasy.ٳܹcooi޺ޔHT %aGc99ikT8y"%%eժU򲶶PSfv["sFO8x= ^Qp&8ʕ/+*{߿аXNE3:9]ǎ}0u4Xxrs=Z)JA!5kVs;}ܹsZZ3Ǎm΄ s=MAapmN Çc5kffW?}*ߘ9B28l_P(SLD+Wp 9oax ̙Y8N8hUW[j(C*>FF_?~]cǴL7oݾOxտezKlm*6$`4`J55/~qذi}###?I-"ԽW ׮%XHքݴYH$ky [EX)K{=QSN8@:tDlȑXX% -+L2kffOә"֋v8Wge,jɱ\.`g'MRPVUU <`5պv}׃Z^_|Uu}6!4҂pϞUTckR ֑(Ξs"AYwX1gEA37LL&6 ?c޷ؙ~xMM^mEQKve*6l]=8Xus'(yU$ZB^^ïjihoLX*쐃m7paXg;K_^f[ajj,bcьz (z5y~A`XC IMM]pAbb_~!Ca^E]̽x`#H$EBMc\R%HЛ7lGFF &lkRt\zR>rhw;e崰nHt>Jzh(}udB{{{ervy"n񅼼/Nn@P|/_ muu TWW8A<8G$Pݦ9vxrE99H.ZH<OݴiE'KnݒK6)gݕ%r:M!Q.>5"C~]k]<]X$ ٶI<9vB1_&PzׯeOʢdW^ 8np-N!4p}.CټWLsHҔG7oSQǓYV?5h-q0԰\\^E6n7G_62T*(*BPߔI&=ydȐ!G9}gYQO3{MbrصkAOn{fc ի8.۷74q١d䀀'뻴(C>.;l;;B#- =,k|;j]Èe&y-q u vmmxz}OeIˆ|>ْ%҃^S!ޥNFF /(1 ާO .<@sDޫи7&\jpߞ=/Hvz[p?׮L& ,11;00PM_3[]ͺ}*r)],,횋06t*M+Jeǭ :5TH* V_Þ=wx)5wEN6,4t uuXYA.>m ;T bc*bA2 rs\Vo#BBBN>Mя{6[ +'.3յk?EG߉ixqעduep̬we˨|~92?R,N\ݵ7R31NZ%g{L{o~: u H"ӦMc6vi]] a>9O (**_B"'Aʵ,;;p)CY9%{7~,^\Щ411 LN}hٲe/^Ѩ+pqM OieevCiwŗJC !C+wƍ6vM7Ǿ Xʥ NZh迅atfai: =0 X,VpM ͍u,+g۔$FQQ@x12y[a~뇩),ږb1ؽ/,DFƼ0hcŭ 5=<;6 q`v|xHprR2,+$J1ܔr?/]JHN7dȐ$46vq9̍}2Uv7j{o4Ե=H&*.I$_mmTȉa͛W^X/q8f%yrsS߾MP ³֕Vq$AiRoC h6mL01ͮL&uV׮]7nH"/_;xpmp>]nՙ%}fh e H8y;fvg?@)ܕѹs792jQ%coܺ:h}(LJaܔ^m\b)28i9;"##t0L|.V~ʷ<ϺhuWw!RRR~qY/qDNNGG9u霘 ^z7X怯yHLLZ[GxګWlaܹsB_%V10']:JWBJ8҃}&脭)-[$dWtaX ޛ'kY`gHI\-*T-B">sT`jbn:F 9ڵ~A4Ttlxׯ<<`tVXcYYB555 .twwp|B655;ǎ-2Dd2T`oo7n=jjj A k낧;Wx*ff}H9#~|UK&l6ڵK24tI\ ::{S7sz*}* R'-g5Ʈzv)FyrTCD"q }Nx">>>##w^XXT*;vlzz7v.e͛`S|Q})Z{U֯_!DuF2 7ήXt*ZVWfUg;{9W話O(t h(a _diRt!q>fKKYeeUNmcccSSScc#p8,K,hck!W,x70:2-߶ҥd˗O~&5!H4o523IF)Iw+}G2`nR/+(4VV**:֭ɓjh)ѣS VNLMiWV#\`Att4jq=J$#F?Ǣ`FuUo8 gbx)uy;))lE(i^@1 7;v={v"" ^[Qm%ct]_DOi4ڒWG߭ǣ'T}BΨ%X[ gdttF4s٘>2ad-Vj8Q~`n ̘ak``gee,aXYYaѣ:IIn܈d05܊*Ce\$*Cm܆7nܢE\(&3oNi]|IjkM!ɡ÷l&@*K*-j-ץ3rҪg,=P,ǰ =u+Rϟkjzfmɔ"H ݻ7o޼;w2M~?~^OCY9GM @N8qٲ2.Oמ.\P''O?4T&E}jm؉IQ#) VqIm>(D-[V^me>Ǣ OvS5^Rd2t7oOXg_7To M0N;v̙D"qf D PYYŋ^z% İ.]ܾn``zmmयi矸&8Bz,o! l$ZZZtuq*mp첳utt+ܕ2P$M06FaÅ =יRx)(9o߾ 1tb>6kppLnjSQQtQ+++U]RNŌ˯x/C9<*Yy{wUru%KbccR)Htrr0aP[JV[M斧v2u:Imɱ܊ӷCs҉HFåF$AWWk;t=Z;ڵ<OKKe\p%O}q @%pC ?*"p\&x6n6MB1_(拥"B8yY1L4nJKR) 7lid Ǯ! x$>xسieLMf||g^|ٳggg*XYsDvիWF_dpB BNEoa>DT@]Ut)t !RTu UUgG$4#du*agL6ELZa۱g)((زeٳg---_ŗq?"h:|gCo6GΜ9S+(>j#N59U{rR ҊsSM_W˨>"@"$-XOph?C]uǒJdž  5k-[`#TkA]ݲ 8B IxBKO~:-#D*B฀^1X8. ނ#YrS.'$th>qĪU444_w7͛P0|rNp%eUl+v++NL\]V`4 p(lh6E׉_GuRt (ٰtqǦ|wtWTBngH$T*y<^IIYQ~'dulh:&g@ B@SSBH$@*]daa99݋N".g/)sg8d>JW>*&M҅ak~lۿwHSRtO qzƊR(492ܴ 艈ά1xǎq/I#7dop4gea...>>>ii?0c2!B:mmn'Zt !d `a՛&\leLT;,$ݝ MFR0d`&! qd1 < fPjQ>xN.N}A$2Δnvuha3Y|Dl0_iJ[#m }{_>jesdaTN,!);i ׯWDMEQE2ӵ2pT X5c A @iRtm!Uc@mOm^ne];T ۷("EF˱ W~ѶF.#ukeg14o']0ձҢn8 ;y>>C[NGס{նC9::bp'*a 5jp[dlF#\ %ǣDBj=Ɍ*sͽda^u6*MzUtb:H$wQUANZEըQ;&>|8tP ~TB:ݫfYlgLMCa|!mĞK,z?N*vɅ'{y| ӉX,~|θ2R|nr)Ug!^SuFpFP,];W1-' O0H;W]]_v3rx4`~\.^O3o'}K Tj+IҸ8EP-U *)J"vR[JUJ*]r)UTcP.epll9C| gGF>ڵkORJ]]}ʕKjwCTT<*#]Mw4K"(V"E "b2:DI&稃ѸQј#c0Dd`@ 4 on^GuL~r֭VgZ܄L&ʹ #D o9oAAA{3b&r'-F[V! s!LNNf)))&&&f5B.\@x/r@H+W>leeс⅃ޱs! bggPJrCL7FRo/RhS剨j kV^=\.P(Dd233k׮?Y3G}4* &i-f#ʹ?2 Z-_@d2׬Yf͚U @ ?xuxjzppАLߥ====dh{ L͵rrrjkkKNN޿!t0RT{ T+rtq|ddL&ϛ7ozӍrohhdNuvv[ZZL&p8?Յ o```b<%:::TtiI.bq\ tvvjiiMڝ$Ȉ ±1r߿/MMM'>@޽;::jffFJewwTV(MMM<266RӕE&wuu*&ꪫE")VT]]]<000 SPvSSӉlkk0lhh`LJEd2||vxb* W8yǎg777[[ۍ7͛7믿r ߿rʄ+V:t_\\pBss;wN} %c||VRAD"Y~֭[###}}}% wwwY>::߶/ C<<ƍ|7n8pڵkX,VttRqcҥqqq|۳g\.[h xAAƍ{=+υ\.x7r9gΜ ݽ{U""" \\\k׮(JU*hhhttի766P(/^L')a&9222 ȈD"$BPA`}I(r===uuuDC( MLL`lddRt2L"d ؘRA hBaCCĆ0--- }```ZZ~2L,CפH:22=!r~~~TP!"sK6bZm6!0ɩEIR(gX8.,XP]]Ν;?aؘuKK mǎEEEІ=z4!v/|||& Fh^OYP/tqΞ۷Looo<< 8. j̙3>B!H$ qb,///$$7tuuNQGGNL3 *m77  foذ_yy9̟?aN-ܐ͉pӃxs><<_{5T*4551LSSSbڤ6a~~~$`\fRiuumϞ0 d2[VV&\[[`0LLL8u5E$544xx ~~~0@^^{֤޽{: ΗvT-,,|!N _P(gϞݼyXvv6䄆>1ȳof`3Lxmll<<>xy*:1g`<~͛7]qqƍ濷W&_Nq[XXLֈͤWBfxwv@WWWXXǃmQQ3000GEDDV𷀵\.xX<::jeeEP>}(33.,P(/aurrqׯƽH$BY`OmmyT5LRNNNK.mnnh4!!!D@ `0kjjD<ͰUu*###7o۷oyIɜloo߾-+**ݻhhhزeKll,/))iiioڴ +))MMMrʊD" ]zL&/\%H$Ǐ KJJ~ǖˆ"ߗ奥REDDdeeUVVVVVfggoݺU"xyy|~PP^iiiUUlBqΝ@AA$kx=266i&ssP ^?#\.Vm*~ԩO>w)|˗/9s&>>dxxD"ƞ={ƍBǧ){Anjj|}FFƳ'***33p8h@FFFuuu[[۵kWZ>|xUmmmAAA{{O>a@DDĿw\~}۶m8[L&ðp8{as8M6^ztt4((nה´~7nnnFFFiii666 %77.+LLLMMM;;;-,,D"QzzׄIffP(o/_z1@jje޽GR\.~:`ҥ, ;00|-$kYd a-j4x=<((jIӣ}].\xJ^SS3qNܹs' fs܉4ޚ:R!ͪbeܹH$G4A^^y \.gY 7aG9re󐒒uV8:/-ZA1 yZBC@ /f"RsLT*>onn9g, ###_O! `F@ڭi9_yfx._|J---'.P(uuuhb1 8;;kiiԴjkkbѣG:::˖-# /Wigb:h+Vwާ~W_޽lŊ C"[L&:33g{{%K.^c#H\.766ѣD;rHEE{ggkkkE"ӧT_~`0ܹsĉǏ.[liiivvv0NJFјLfoo.DO훛,Xp6scccǎD~bG# @<ؔ^oooHki4?III%''ٳt޽{rs}}=aJb566¡Ѳ2wwI{pe25qajZ$ݽ{ðF>~Iqfdd30 ;p@HH$JGGG1 [nƍsrruxxx^^ޯ<㓗ggggnnrl;Nljj B}}}>ݝ eoooyggIx<<<#""97oLII144TPhcc3O [__vZ8*L";>>N-[FJΝ/?nii ~9@F!T֭[w!+++???}||y<q\WW7..6L& B:SԞKK˨/ׯ_g0o?wp8Gv?-aMHH{yKx INb(éS? F]t )p@V SfX$, 3e˖CCC:::\.399D"QZE-LNLL\z֭[R)F)++[jP(R F 8hjs0 ##b022fff H$ h4Q,--B&Y__/ÓT*FkkkxӔL^BRwfaaʕ+WZ kMPPT*J ...W^_vZtҥy222:{Çդn,k޽$)%%>H$~ixx8N_IIIYYY.\Xd Xaaaaaa<gee^722qѣIIIpaɓ'0 9wz8 ~#-_ ?K7qy?ϊZ=M s 2ә_K"J`j ?_ ~0 @V' endstream endobj 385 0 obj << /Type /Annot /Subtype /Link /A 386 0 R /Border [0 0 0] /H /I /Rect [ 380.6723 214.3284 390.3096 223.1467 ] >> endobj 386 0 obj << /Type /Action >> endobj 387 0 obj << /Type /Annot /Subtype /Link /A 388 0 R /Border [0 0 0] /H /I /Rect [ 35.2500 428.7592 485.2500 621.5092 ] >> endobj 388 0 obj << /Type /Action /S /URI /URI (http://currents.plos.org/outbreaks/files/2014/10/nowcast.png) >> endobj 389 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 390 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 389 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 42823>> stream xw\Ggw:DQ,HQFb4jby%c1D_[)jX[DEPAPQ(w\z11&;|vggi`A@ +${+@ =A@ dѮA@  ,@  !@ 3E@ v j"]ZhנYhא!$H"d@  E DBDBVWWfSR) hrARi%gϞ@Dɓ'NܲeˤI۶mYvm+$-'Lpqȑ#={tuum=M|>`۶m&M RcO ʕGf2fl6ZV FTTJUZbTTTSTLF"*//DBC$t:K. `6 _HHHuu͛7drppի|>AwƒdduLhhh Ph4l.--!!!~~~FQ?~8!!D"ݿ_*FDD[(++^^^aaa4MVL`R͛7T*Upp0D.ƲX,@eeWqq{HHN+**tH$z!,&l h4\.:vhU,+44>px䉏ϭ[\nTTUrөS'?BPє(.Ι3zjС]v&);;tѢE.\(..NOOw333322ϟڵnذ`L2%,,F*8qfO6!,,l'NP*Vp8_}Ո#QTTԻwu֙Lcj Ry,̟?q*zҥ;wSӍ9D">}۶m ^xqAAAtt}>}[.;;W^"h}IIIܹH$:x5/d[u6o)ziii!!!^^^ hF X,"رc<oΝ[lӧύ7Fm[ "---::Hҽ{j41c0LCOGK,ϟ8qbjj^ _o&%%ݻw/<<믿q{ɞ999o;s秦T.]̞=ū=#m~aȑ?H$+W3go%Kp7C ٽ{d _n]FFƼy,KǎE"ŋ Cu:R|W9( rj]]]@@gLL̽{p>v B (w0؆ zlFL&Rǎ5kƍgX={B%oݺաCJ84~O>W\/,,ѣdzUTT㬬O>qU\\x@@aI&m޼Ϝ9s֭2 񆆆 Fc}^oF,cvʹ4 lCQT Êl6aIII=RSN?~|VVf3l͞=͛-7ۘ|>_.h4Z\\z~aСC Ÿq:~x\^]]m-22o߾ ca {)//>|O>ҥKaGyѤe˖¾SLfϞ=alnnn!!!766QFm߾}+VʧcH"v$۷oVVO?%$$>}a>>>_~=B5jԾ}l:t0k֬4ww1cƔhѢ7xh4}wBКkqJ[nׯ0̙3> 44/T( .lhhR -իW_~˖-틉w$f"Ç7|>L&}ѦM(JBBBRR5~ZmM?СC4ĉu:… ׬Y3fg?~FWW۷VZ{ۗݻ6 T*_~$dܹfom 5I'bM,wڵ޺ Xq={g}fo?9ٳgȐ!>>>VᴬY&??`0teN@Eh4uyݻN@+`G&a0C+.ӧ[sss-Kbb"aR4==}ذaV ڜ:ujС᜜44رccƌh۷ogggwd25Qٳ*jΝ...O?{ȑx߂t:]UUӧ?~ g՚ܻwoذaݻwoղCj"|Q^^^w_ pD|&IĢEf̘N>=~ \޿R 7o90 .4Ldڴiׯ ۽{_~ uVfff@@@}}~D"FRw}w̙\.~Ɂo~MJh4*׮]UUU>>>hI@-BsioMyW6/ ///p8/S[[O< ;w.y}-[^#֭[G8qs#0l˖-t:O?]hѷ~;s̉'Ѥ_rǏ bYr%t{&;u$  4h+W\re̘1/_>BJdffԼ|xzVԥK3sѣGCChm)VVVFFF0 xɟEZ...XYY 11 ?d "2qr7{7nd:tСʨ( ***_ϯ{]]]m$;V9jUk愻::"K~'hG&LHKK5kP(Dzxb:ѣk׮__8>L[&O>>Ph/ 9 8rn̜9s9„h!~Z Zy-[wW.^F8'h@ "z^:d1%i@hf@ / 2D3#4t/Zj@ "}BkPMz$@ NhM&Vuٰ.u Am0#!hDžAt?0 Oǩ bY~}tttllQDc0-ـZT*6mxҥK[-/>L&ʂ}8ggg3&--mԩpӧOufڲeK]]hi4 6ddd <?Pd2|-^غ8!=zҥ]͞oZ0K,r @W!l5T*?`F@}}}vvvVTT<{믿NNNk.]tI&m޼yȑ7nI8ݻwFqذayyy;vcǎD"ĉ?Ss26mOJX,@=G򉲲>lÆ ťG;w=Ię3g a,gΜIAf:q //Z9soׯ_ZjРAIII}Sƍӧï]vȑ,Zhر"6W^D7߿РACa0@f͚{o޼yA}G}m۶:@ddd@@#""NJ^رchh*l&6W^yjo`0,={wL6b>?k4m۶M:…,[l۷x<aqqq{Ν_x_Z-$mQ8 NeZ{=FӥKWVrd-T* 5@-?!޹~y3LHZ~SNeffB .\n_x?… f~2eʃjjj,X@ܞް~I*3}twʕ]X,FjRUU'tV\_N1 *jjuHH B>|x޼yBP(=ĉqqqs Ϟ=[XWPv~s;u1FQ.j4p'$I p"##?~񂃃-kdk}=L&BZPSSCF"(%`_ ؙV2o=Hc+?^zY8JP*A<7߬>uTcccxx7x<F} 7o^bb"B)**[nϞ=+W,--ݼyA`ZMhpw' +++F#JmR> MU]] crD&mTɓm֯_Ȑxh?\5Ilrt9tP?T*#i !277 Z!-"4 r\kP=f{J*Ơ^Ferƍ/H8zjTTaW^͍OMM]reϞ=1 YTTԡCӧ'''k4xfϚ5~!P(FVaaaD`ַo_'ŶT*'|2vX 2f̘'N0z eddk0 #"//sMbkaǏ?y_~9m4 `Z7Қ δ}<]O?tܸq/^oVllX,RBX,nUѬSUڵo>}СC777@NSSSN:}t1|p\X^^>o|cǎeeeF̙3A={633SVWVVZ ><11ƍ ㏭y\`ȑ#ѣX9vfpݺu[SB___~E"0 77(!܀ vjjj44{G!ls౗ו+W`$9}t^^V<<<`úXXXرcGx~ƌnnn=z`2999ׯ_ k3V^]TThF |ƍS*Z͜9s޽p&...//P*;6220iҤAad2]]])SdddLwy'886N:]rڵk W^L&Tӧ9agϞWnnnSLS9s&11>l9TjkDnl[FJH&Q f4_5QQQ֟:t3&ჂaNT9&ƃBئheС3Lp'۷mgۻfYVqH$(),D;v~DK,AM+fjkjj.ˍB w*h"{kAlomG " RWT= k-%DoozZuʳ aܹslVjZE"*+LfJ!QRBQ@dgg!h4Z688TV>1l BGc !-  .['C2kp!D 2 qWWW3SI?%ku }UjCWN'HF5UU* ٫A<;YiF)//71A!D8'y>zyy N0ssǞAAAK.͛M(_|s,X@" bƌC޽;.q˖-AAAaaa]t1bmOݻw ޫW&J8nk ڱcG{̘1/P۷oȑ#7oܾ}KfBL8Ѻv9@خPnQB@v11Vl6+={@Fnd,qC1 fX۵Lj`%X,uݩh*++atرNפW_M|xCCBӧl&"..\8|0t`v^zM:uذaG3fǏOMM;'OӧϬY222;vsfff'|ܹSPPfeΜ90a$8pÆ +W$b޽111SN!!!ǏWyy՜h4Nhc*:}t|?--WZ <x~~[x0;o"ٳΝe03g\rի9NAA#.\X`3g Ä vڕDV%b۶mp?|ӦM~֭[׬Y3p@ Ne˖hfÍ2w#sNXXl|vP(d2٭[?{k=f[h06h/^HPD"۷0 h4T*522r^^^s΅q=z'ի}˗l>CjHt9AX,Vl6>{ni i ٚrǎׯ[4-8FGԐ-Wo44 yCd -OKKHxn0X/9s.]:c x>77wѣ{= RRR~&L ENNΊ+JJJlvbb"fCS H.]`=..ڵk)))MJ&-[hllu4l0]0a Nd2_۾m`UKh j*$P. FxNNӱt_K{뭷lْlwFLMM]t)=zw^S[[ۤܿO>9~xppÇ9B 믿Κ5k{* RtqqyAnݺޥKgsz+Vi&N6euĉ>}ׯ_:ujM4 ON6 M`0FWW϶ sj`&>ΣNMV(M#R`0 ٜXtE֬YĉZ'%%aիW'&&bi0.^Hbh4AAA0NX6cƌٳg`p~V:u(0 5 *ojy:ZҥKE0 0aBAAw}g2p>x aT*`0 2Lm!`6dsժU)))111999!!!pR)Ld2dXRRRn޼߭[7>="x^rWƻKH[֔śӷtZR2 ___ S60KVi8ۆD6Q>--!BǣGԩ@ bRPHSTJ/^ypq-!U'M #f 뉄c2AM c/ٰajgl6߹s[nVMas8NeBpܸqaZ!1BAeyP&k9 ҧ&z=P(0D"еk+hgOTSN* g5xI[%h2F=iZ 4[B!NVP Scbb.]{7|ׯKT*m]@kԠdXI2 ۷!t`pWϮPlTUUݾ}oT-Bp?l0d}Y+QI$zYIjkkqww꟮## H V\io-Æ sD+ }GKHRQQa aw}l}ivESJl6b.WXhmP*ZT_t!c]1Zm]])!mA$/hFϚ5kÆ RC -d˖~8rbfBFq>jZR[%8 YWG. X, 騲KJNϥvפmߒN5:a„Doo9s?ugZzVh4\.N-F£t(h4n}uuSjF\^<:%p `\^-{$Q"}̀SM 6AFFFK'wԩa&A! (ŗ ZW7Kʢ V[Katr@I dfm AVt\X"Thdvip1BMt>hZ-\D7jJ6߇"555\.DK$b!$kʃ}"Jn}-v['n]+;{_m>M$-hҩffBX,Vjz7 12U/ DAN3,UPUK  f^7j_۝rWjB4)wS"V.]'[?j7!,&HO7`"S5/'\[]o#Nb $m-r G40B`M dΉ\.o$lWӫ4:-B`+%Q *3* ^ERDA( $HcGv CEcŶWYiyD<}zyyh4ZLl//ExdpROT&(B3cO>C,vyt :a.,w^e)^^]g=фnݺa֢5a\aӹ yyy08JLtrC&Q8 ^w#ȳ[7T32jo{m۶?vssSt*FYtBkziԩ :Nrooz'5?޵ka): 5n6@oڴu]7n(7~{ZY(28+u"gT*0`d/zѯ*hNeްaFޯOW:qɓ'NS(#G,**:wAذT>.l%ZJA`-be~A ,8rȚ5k222Hee~wbDNAtܑA:ww;۹+ˊ.Nr^E@=.U<rh@C[j Cӧ&O&MdwXJmгE-Bswߍ76/--DDDQ]m=[TFr:9`hѢ'N\_ҹ\˦s"ZiDh49rd̘1ѠБu7o$ml)aRe@-BsrcǶܣ/--R!!! 5xHi) IaXVHHH߿Kˋ"oR¼Q--1z+%-Ia5. XUC\Re@;#鴲垾l.++ J?ِt*`T:9j=a؄ n&:eK ɶ&-B@m-AR  jhSCڂ&//Ivנ٥^s观俗K /lJeQ6ֵFeLz  .jZZ=vXtE"@rC/%a uҪA~~nݺq- Au5o %-h$ 4qFy[ ֔k֬!-X,V*5ȍ=5ПlH:i0։0֫wwuD `۽ܐKb/_:iǵP@ڵ+0kld>v/mA%Ivנy\.6]+$LZg*..oJ:a0Aլ,ٜG*5j-6" CCC@ gjk ڵkW ..*Ψ5Y mAgmAN5kT,oٲe޽LyD!$ϭct*d1(2A2yWf̘q}>ڿۛ'i{p}Z)6O]]Ap(Dm-!Poݺ8z6BWjd<_{{:E}l6dAAٳg+DnFJ;( ,WK&hr+11֭[(RьLsεBB)ARp~T77J34w%gϞ-(( =qhLHHPTA(J8:::99?n z":ɓD #(( ~熤S$WK .ˬ,s]]ѯ^!{!h4/j:#h^$ úvT*=zaPjT6c&''GGGtz`04&Ν;>|ZNMM;;S(*v"anfsyyyXXLTT熤SY™2φ2d<~(딂z_WJuCrIAt PXXHPaQT jӌ1ZAD<,k?ۿ#i:BOѣGFNqPScD2̸Y@1φ uנqK[(=$_ [^=+m[b+&cf;q]O3JJ[/qi 4-55ַGU`cuV_qznH:i2u}wƆaXffKN&yRUKH6gBZ-BM` K. ??记TڽdM^^:dgꫯ r!k >i2g)))̔6X7W6Ph^M{ȑ$: 2{YWjk g2FN@-h*lo0Yz}~~@;ݻaXxx8NYF ԛP(ٳgM͍lX,p8pUtHk׮F~ Z7j???1+![d|쵮.s6zyh縕Ne1^ȑ# BS[[jT ` bn9i4e2p8 }s B!P &ިm i tYZVJ^wAڂHܽ{7\>Ѽ1j>FDUOH0fB9p@$5۽|?CKVU*BQBiibr)U:e[(H[c ѱc&'-u^(ۏfW^p6V;!t*FaОYMoKxxxPPPumu N5 ^3p4T*ZeE}WCCC9tD"?b70>{F{ aqڌf8SF҆-?Net:=99yǎ5 &BT*5LЭLM-Neh\AÇ>>< tJz@di?٘7(((yD4# ,>ьXN`RrED4 R;>sX~R$мj#&9{~.",t&J&cbb~駪;xԃ:w/)#[T^t >}<{><<|Ν/s{ (h/fph}zӶNe2hT#55B!UAƶeH$R@@@K?#Q DbJ&ڵkv Pb28"\dɳ'y<(yD}JKK1 0L$\.,TX77=#srz TcMMzغu1cBRU1thKN%(WE4hsϿGD3RQQі BCDv_P}"992//jiphTlʈDhI۷#;f͚  %zā7'tcz/*0 sj`&]b4op( Ĭf<-- DKިW%NDKt* 0oD"mܸQ ,_fm-a6>BϐK'sW{A Z~gUQl󮣢B$p -ᾩKMM Ljl$Zs&' 7tիg H.#T])=sy;^\[Q~d֭[Ν;DĶmG BaXZRH+f#RSSMPb@j<ܞuȑoFUՍ;wJd2ieCٜcFa{uoR[9 c# Gxٳg[#D[d„ ;4 tq#_,aaбWCBMzz? 忦u}#G4t:}ر-bVU41 [v+W?>kw>lhL%;͎_[ h3YZhB_Y{;C 7oBezѼ`01낂dAp+/-aʟ.~eQی# ҽ{MB"f`s.믿ưon&ݻwIIO?Hš*Qsp CX]]=eʔJ\. ܹs\\D Z*۬YP Bmh޹y>>>117nԺ<͇1ra2 9RΟB JKN޳v[2+WtORh^O8dh,8aN;lذk׮z>---33sԩV F;fk@;?{vZj$}Z,ǎf KQѤo E&a5f3QYeV1h/ð,Tʺudee9r6L$&R*8!$$$>|`꫙ϟxȑw&h,]biΙr999xS]붜P4/drڵk;xu1 %U?ߣV\٢IXNT@`_҃ĉ$l݊/ZL^|l^R]NoD;?tׯwoٷoٳ%B]l\OTLJ" N^ {>xŇC("&SsNzkP"*Ϗ L0BaFZG !1 j8!X,YYYvھ}Y6m4yiӦmٲު! vâsZ(@гgg+2J f:"Z_wOŖ,!Z2)a֑BL9Jemc:88xǎ;wtww GaV F2dH3q r0(YX,lf`[bq{uxo-@Rcsq}&)Un [Pߋvgg33;3pWWWkkk(ԩSv{xx+,,Lէ=m+ϟ4S r H{qm̭HgTU"z ¼<}}}==25f~++B G&7 Sj}>*B"xƍÇ9իvرcGTTgϞ]XX}!0 :Cda N_$=!LE555555 XBgֶE(p¼+iS j*Mn#^">}};SN\.[:TI$۪K!==N7#SIzb1`hg7J۵k_] ΫRK_/|H" Fa|vl˖$OؑDqz~;MEB/#Fz(-]hĉwahh pww߿^s/ΦSEE $䲲7"|<ɸ~0ɩQZ_zׯ_..^dfH"Lum|ĤmT;Bd*ڑJ w/)|!@] mmHݻw[h7 To6㙙VVVt:WcV3"0ld۵KVSQhmX\zm:XElٲQFyzz:::8PюTRLB }nPZ.!jfH-1h ,Ԏ`v\,FZt=U.oP!J655m󏲪}ŋgnX[ ovj)/_ $./xuC8u,^7sR"dA`kVΜ9SKK̙-XA\&VEmPcj۶m_>}!`0!Ih48ofgMv%52<884O8 76I ٳ+E!tvv.-EDc]öYCCc,Ν Jx"yEiWU*uL~ϛfYZZ"G 5jk*i݇S$Wf:]qUxYY׈!ebbRR^eӚ9sڑ#Gtu%sb~ UJ*}bwL&C,ӳ#f608s&96l7sID_RC̉6ѱjyҗᔖ0̢m+B;;ƺ@@]]O;U꘲YYcTU^^^ՈJg}ˑ2KC]k=?<<]6Y}Fի8_\YZZ~+**X,3aBF&] .\!tA heR1 =6MNUZ|gO(/G HNF oo@܉@X?k$}hmakccUQZn+JMMmѢE_#LTKvm6RY...EnnXJ neXRw*BTPF xBBJuq\Sj}%Ktf zX)#1o_LJ)|&:Ŋҗ(33RQD+0 [`L&;rH.;rXz"TI'Go_&)sssH<9t"*[/Scǎ,_r)..>n>C'IqGSwU9H5=*==йsLD֙ehƌc``p9]]TCWmM8W{"Tcɓ7}p#8x<`0X}o<]*]믙^K׷c}McLP6P([ɓ'٥v鹹AG?H۷oɐYS]^_hV2,% aU-ϡk6à" \]й2NPU .|&ImDUYL++!tկQi\ B&R`g-VxNu2w:bM-Yv0ҲPUNo޼dNN.׮v)Z4k@ <8555)tYSw1׳L"Vf*B !}拓{./ʜsdHdV] H:2J&R-555*jzI,<`ɓ'%'J>2T 6XWKhhpkGwH$,ڶmܷg<#3T}^6+WT^^NFVll%K2t@6h_MAAAlL *禤lx#Az~Cm6y25~jjBcs{Y١Z!CLXK$\.wΜ9nݺpBd9{S7A/+C^!,/ZiCT] (z½O c`Ѩ88r0ϯ::-! XZebK(bbb0-]='CyϟX[ΝR)0׳ r xyZ>KZby[RH5o-t%:rG[nرݻI| L_ :ҒP,`=+aǣD"sNƙyf ]A }YZ7mZ.y`<uϋS.Qk)W"p|*>ܽ@ԶgLMMݺ5Сyx+!:"hѢɏ;_6Iu,&䖧ꮭ޽{gىbx+T^qZF$B"^ݡpu{C!H?<#)/RÇm[ccٳ%fѽB0-S4ͨӧOhTF،HKJ& !؈lo.Ŗ- ̘qK1ْ%ӥsHGKX6!&p#^paӭ[hҤA}z.ݘ?$V9 t:ײ&Or#ݻwpΏ+*LfkDIFUS)XlF @ɓ'iixD/I w'ڰaCJJ6SgLoH$nBX(/!n@"R^F^r)kaaa7oNH–.LM*޼ysرF9, w6r>'|J.]޾} ,RIm2XG\- Xx54Y_e4 nC "lB54 CC1c,tK bep{SSay,o{wXgN#u Ͼ(~Flr)}^iKz8*0E:RWw,4p>n!Eq5ܲH2QءZ{ݲeKYY9`Xl4 C8p`ee<<aFFw.\uuAZiiiFJgΜ3f tzHHW^YH$"!deeٳ*ȩɤ O>H${Pm"ӫBܹs5T"7 [ھTΟPu*e𣢜%%%-*,, y z޼^n]C-immc:tgSbĴ705Ln\ʐ-gjΟ??**jƍW^q<66`g!N***ϟo 23[񼜗<@fd!lhh?~<޿s[cf3]-!9r$aaani[<0_Rw<`EZMZt}ui 8t; kmLOP{At$J҄;͛MMMk׮z]di`qUN2elo{F׹%ȥ𴴊EH !LtttEа}v B&ϟ?/᝝7H&kQ2:vL@"̝;7++k޼y3f̀=͉ׅNEo;rwSqMߙvƮ6lLv횋  KnE:ǏkhhxyyՁ&aFe ԩ &]U(2o6I'Ovq=~e HDq,BM4 B!=l썻(CB?~|ҤAMM;ʥ'Gjաun6gpW3e&M %SЕo~*o>qq? ؟6n܄t}iR<,88XPTu 8k׮={˷DDD ?"D"fSS<ɓ͛3!A6#P($L&GEE łqN=2\Rޓ[/OoCR.O,gϞca!!9lɁT_I{WiiӚ7oÇ:e޲f1ͽ)*J6y``7ԩS%uIҲ%LfUUUCoHѯ{/JᰭTPwj;E{vDDDDDD|3m"bjӗCv6bPU02RrQZQTRuà9 aT\àw: \Q׮V&H B$KB'NJ%cG?ihΜq"õKpə\LiiiA?>J.;o2""d5oJ %;;[e"۲Hٳ%gp7nqyoquζkol|OLre:jV|IlVIO d^^2uVkkAe. ̽W.LT(L1r2\_F|@ x?dsר=p&˖-ϣo!'O3-EG?86& k%&&7n͚5><ʳC=Logj ƍsrvʮDbs>(B ]] )>}mDGG!9fF4յ!TSyӃ0/etJxmGS8{k tx,, ! weXGeIU=m=]Q׮] zjz:zZ$''7СC%%%pe*vwZ2龍\ LMElat41qmDĥݻ=z}!t] ޮ# v;ursv4Ħ|`"[;\W*n޼944N)"e2m8QTVSʐJ5Zf ZnudDpoxqMީ$oر~TfaJ1I_[KkkO[{İ?8{)S/XȈdN<;eo~k`R_9!mlM2]0:Z0'g.]^`0ͫL-O9mMc*_OYL/.N3}F<xzr |Lʯ8cƌuefrqQ[e4 %鎎ܸϵ9>Ojd$+[$\oasy.ٳܹcooi޺ޔHT %aGc99ikT8y"%%eժU򲶶PSfv["sFO8x= ^Qp&8ʕ/+*{߿аXNE3:9]ǎ}0u4Xxrs=Z)JA!5kVs;}ܹsZZ3Ǎm΄ s=MAapmN Çc5kffW?}*ߘ9B28l_P(SLD+Wp 9oax ̙Y8N8hUW[j(C*>FF_?~]cǴL7oݾOxտezKlm*6$`4`J55/~qذi}###?I-"ԽW ׮%XHքݴYH$ky [EX)K{=QSN8@:tDlȑXX% -+L2kffOә"֋v8Wge,jɱ\.`g'MRPVUU <`5պv}׃Z^_|Uu}6!4҂pϞUTckR ֑(Ξs"AYwX1gEA37LL&6 ?c޷ؙ~xMM^mEQKve*6l]=8Xus'(yU$ZB^^ïjihoLX*쐃m7paXg;K_^f[ajj,bcьz (z5y~A`XC IMM]pAbb_~!Ca^E]̽x`#H$EBMc\R%HЛ7lGFF &lkRt\zR>rhw;e崰nHt>Jzh(}udB{{{ervy"n񅼼/Nn@P|/_ muu TWW8A<8G$Pݦ9vxrE99H.ZH<OݴiE'KnݒK6)gݕ%r:M!Q.>5"C~]k]<]X$ ٶI<9vB1_&PzׯeOʢdW^ 8np-N!4p}.CټWLsHҔG7oSQǓYV?5h-q0԰\\^E6n7G_62T*(*BPߔI&=ydȐ!G9}gYQO3{MbrصkAOn{fc ի8.۷74q١d䀀'뻴(C>.;l;;B#- =,k|;j]Èe&y-q u vmmxz}OeIˆ|>ْ%҃^S!ޥNFF /(1 ާO .<@sDޫи7&\jpߞ=/Hvz[p?׮L& ,11;00PM_3[]ͺ}*r)],,횋06t*M+Jeǭ :5TH* V_Þ=wx)5wEN6,4t uuXYA.>m ;T bc*bA2 rs\Vo#BBBN>Mя{6[ +'.3յk?EG߉ixqעduep̬we˨|~92?R,N\ݵ7R31NZ%g{L{o~: u H"ӦMc6vi]] a>9O (**_B"'Aʵ,;;p)CY9%{7~,^\Щ411 LN}hٲe/^Ѩ+pqM OieevCiwŗJC !C+wƍ6vM7Ǿ Xʥ NZh迅atfai: =0 X,VpM ͍u,+g۔$FQQ@x12y[a~뇩),ږb1ؽ/,DFƼ0hcŭ 5=<;6 q`v|xHprR2,+$J1ܔr?/]JHN7dȐ$46vq9̍}2Uv7j{o4Ե=H&*.I$_mmTȉa͛W^X/q8f%yrsS߾MP ³֕Vq$AiRoC h6mL01ͮL&uV׮]7nH"/_;xpmp>]nՙ%}fh e H8y;fvg?@)ܕѹs792jQ%coܺ:h}(LJaܔ^m\b)28i9;"##t0L|.V~ʷ<ϺhuWw!RRR~qY/qDNNGG9u霘 ^z7X怯yHLLZ[GxګWlaܹsB_%V10']:JWBJ8҃}&脭)-[$dWtaX ޛ'kY`gHI\-*T-B">sT`jbn:F 9ڵ~A4Ttlxׯ<<`tVXcYYB555 .twwp|B655;ǎ-2Dd2T`oo7n=jjj A k낧;Wx*ff}H9#~|UK&l6ڵK24tI\ ::{S7sz*}* R'-g5Ʈzv)FyrTCD"q }Nx">>>##w^XXT*;vlzz7v.e͛`S|Q})Z{U֯_!DuF2 7ήXt*ZVWfUg;{9W話O(t h(a _diRt!q>fKKYeeUNmcccSSScc#p8,K,hck!W,x70:2-߶ҥd˗O~&5!H4o523IF)Iw+}G2`nR/+(4VV**:֭ɓjh)ѣS VNLMiWV#\`Att4jq=J$#F?Ǣ`FuUo8 gbx)uy;))lE(i^@1 7;v={v"" ^[Qm%ct]_DOi4ڒWG߭ǣ'T}BΨ%X[ gdttF4s٘>2ad-Vj8Q~`n ̘ak``gee,aXYYaѣ:IIn܈d05܊*Ce\$*Cm܆7nܢE\(&3oNi]|IjkM!ɡ÷l&@*K*-j-ץ3rҪg,=P,ǰ =u+Rϟkjzfmɔ"H ݻ7o޼;w2M~?~^OCY9GM @N8qٲ2.Oמ.\P''O?4T&E}jm؉IQ#) VqIm>(D-[V^me>Ǣ OvS5^Rd2t7oOXg_7To M0N;v̙D"qf D PYYŋ^z% İ.]ܾn``zmmयi矸&8Bz,o! l$ZZZtuq*mp첳utt+ܕ2P$M06FaÅ =יRx)(9o߾ 1tb>6kppLnjSQQtQ+++U]RNŌ˯x/C9<*Yy{wUru%KbccR)Htrr0aP[JV[M斧v2u:Imɱ܊ӷCs҉HFåF$AWWk;t=Z;ڵ<OKKe\p%O}q @%pC ?*"p\&x6n6MB1_(拥"B8yY1L4nJKR) 7lid Ǯ! x$>xسieLMf||g^|ٳggg*XYsDvիWF_dpB BNEoa>DT@]Ut)t !RTu UUgG$4#du*agL6ELZa۱g)((زeٳg---_ŗq?"h:|gCo6GΜ9S+(>j#N59U{rR ҊsSM_W˨>"@"$-XOph?C]uǒJdž  5k-[`#TkA]ݲ 8B IxBKO~:-#D*B฀^1X8. ނ#YrS.'$th>qĪU444_w7͛P0|rNp%eUl+v++NL\]V`4 p(lh6E׉_GuRt (ٰtqǦ|wtWTBngH$T*y<^IIYQ~'dulh:&g@ B@SSBH$@*]daa99݋N".g/)sg8d>JW>*&M҅ak~lۿwHSRtO qzƊR(492ܴ 艈ά1xǎq/I#7dop4gea...>>>ii?0c2!B:mmn'Zt !d `a՛&\leLT;,$ݝ MFR0d`&! qd1 < fPjQ>xN.N}A$2Δnvuha3Y|Dl0_iJ[#m }{_>jesdaTN,!);i ׯWDMEQE2ӵ2pT X5c A @iRtm!Uc@mOm^ne];T ۷("EF˱ W~ѶF.#ukeg14o']0ձҢn8 ;y>>C[NGס{նC9::bp'*a 5jp[dlF#\ %ǣDBj=Ɍ*sͽda^u6*MzUtb:H$wQUANZEըQ;&>|8tP ~TB:ݫfYlgLMCa|!mĞK,z?N*vɅ'{y| ӉX,~|θ2R|nr)Ug!^SuFpFP,];W1-' O0H;W]]_v3rx4`~\.^O3o'}K Tj+IҸ8EP-U *)J"vR[JUJ*]r)UTcP.epll9C| gGF>ڵkORJ]]}ʕKjwCTT<*#]Mw4K"(V"E "b2:DI&稃ѸQј#c0Dd`@ 4 on^GuL~r֭VgZ܄L&ʹ #D o9oAAA{3b&r'-F[V! s!LNNf)))&&&f5B.\@x/r@H+W>leeс⅃ޱs! bggPJrCL7FRo/RhS剨j kV^=\.P(Dd233k׮?Y3G}4* &i-f#ʹ?2 Z-_@d2׬Yf͚U @ ?xuxjzppАLߥ====dh{ L͵rrrjkkKNN޿!t0RT{ T+rtq|ddL&ϛ7ozӍrohhdNuvv[ZZL&p8?Յ o```b<%:::TtiI.bq\ tvvjiiMڝ$Ȉ ±1r߿/MMM'>@޽;::jffFJewwTV(MMM<266RӕE&wuu*&ꪫE")VT]]]<000 SPvSSӉlkk0lhh`LJEd2||vxb* W8yǎg777[[ۍ7͛7믿r ߿rʄ+V:t_\\pBss;wN} %c||VRAD"Y~֭[###}}}% wwwY>::߶/ C<<ƍ|7n8pڵkX,VttRqcҥqqq|۳g\.[h xAAƍ{=+υ\.x7r9gΜ ݽ{U""" \\\k׮(JU*hhhttի766P(/^L')a&9222 ȈD"$BPA`}I(r===uuuDC( MLL`lddRt2L"d ؘRA hBaCCĆ0--- }```ZZ~2L,CפH:22=!r~~~TP!"sK6bZm6!0ɩEIR(gX8.,XP]]Ν;?aؘuKK mǎEEEІ=z4!v/|||& Fh^OYP/tqΞ۷Looo<< 8. j̙3>B!H$ qb,///$$7tuuNQGGNL3 *m77  foذ_yy9̟?aN-ܐ͉pӃxs><<_{5T*4551LSSSbڤ6a~~~$`\fRiuumϞ0 d2[VV&\[[`0LLL8u5E$544xx ~~~0@^^{֤޽{: ΗvT-,,|!N _P(gϞݼyXvv6䄆>1ȳof`3Lxmll<<>xy*:1g`<~͛7]qqƍ濷W&_Nq[XXLֈͤWBfxwv@WWWXXǃmQQ3000GEDDV𷀵\.xX<::jeeEP>}(33.,P(/aurrqׯƽH$BY`OmmyT5LRNNNK.mnnh4!!!D@ `0kjjD<ͰUu*###7o۷oyIɜloo߾-+**ݻhhhزeKll,/))iiioڴ +))MMMrʊD" ]zL&/\%H$Ǐ KJJ~ǖˆ"ߗ奥REDDdeeUVVVVVfggoݺU"xyy|~PP^iiiUUlBqΝ@AA$kx=266i&ssP ^?#\.Vm*~ԩO>w)|˗/9s&>>dxxD"ƞ={ƍBǧ){Anjj|}FFƳ'***33p8h@FFFuuu[[۵kWZ>|xUmmmAAA{{O>a@DDĿw\~}۶m8[L&ðp8{as8M6^ztt4((nה´~7nnnFFFiii666 %77.+LLLMMM;;;-,,D"QzzׄIffP(o/_z1@jje޽GR\.~:`ҥ, ;00|-$kYd a-j4x=<((jIӣ}].\xJ^SS3qNܹs' fs܉4ޚ:R!ͪbeܹH$G4A^^y \.gY 7aG9re󐒒uV8:/-ZA1 yZBC@ /f"RsLT*>onn9g, ###_O! `F@ڭi9_yfx._|J---'.P(uuuhb1 8;;kiiԴjkkbѣG:::˖-# /Wigb:h+Vwާ~W_޽lŊ C"[L&:33g{{%K.^c#H\.766ѣD;rHEE{ggkkkE"ӧT_~`0ܹsĉǏ.[liiivvv0NJFјLfoo.DO훛,Xp6scccǎD~bG# @<ؔ^oooHki4?III%''ٳt޽{rs}}=aJb566¡Ѳ2wwI{pe25qajZ$ݽ{ðF>~Iqfdd30 ;p@HH$JGGG1 [nƍsrruxxx^^ޯ<㓗ggggnnrl;Nljj B}}}>ݝ eoooyggIx<<<#""97oLII144TPhcc3O [__vZ8*L";>>N-[FJΝ/?nii ~9@F!T֭[w!+++???}||y<q\WW7..6L& B:SԞKK˨/ׯ_g0o?wp8Gv?-aMHH{yKx INb(éS? F]t )p@V SfX$, 3e˖CCC:::\.399D"QZE-LNLL\z֭[R)F)++[jP(R F 8hjs0 ##b022fff H$ h4Q,--B&Y__/ÓT*FkkkxӔL^BRwfaaʕ+WZ kMPPT*J ...W^_vZtҥy222:{Çդn,k޽$)%%>H$~ixx8N_IIIYYY.\Xd Xaaaaaa<gee^722qѣIIIpaɓ'0 9wz8 ~#-_ ?K7qy?ϊZ=M s 2ә_K"J`j ?_ ~0 @V' endstream endobj 391 0 obj << /Type /Annot /Subtype /Link /A 392 0 R /Border [0 0 0] /H /I /Rect [ 380.6723 214.3284 390.3096 223.1467 ] >> endobj 392 0 obj << /Type /Action >> endobj 393 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 1 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceGray /BitsPerComponent 8 /Length 671>> stream xA 0 =" ʝ<2;EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EX$EbYL endstream endobj 394 0 obj << /Type /XObject /Subtype /Image /Width 600 /Height 257 /SMask 393 0 R /Filter /FlateDecode /DecodeParms << /Predictor 15 /Colors 3 /Columns 600 /BitsPerComponent 8>> /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length 42823>> stream xw\Ggw:DQ,HQFb4jby%c1D_[)jX[DEPAPQ(w\z11&;|vggi`A@ +${+@ =A@ dѮA@  ,@  !@ 3E@ v j"]ZhנYhא!$H"d@  E DBDBVWWfSR) hrARi%gϞ@Dɓ'NܲeˤI۶mYvm+$-'Lpqȑ#={tuum=M|>`۶m&M RcO ʕGf2fl6ZV FTTJUZbTTTSTLF"*//DBC$t:K. `6 _HHHuu͛7drppի|>AwƒdduLhhh Ph4l.--!!!~~~FQ?~8!!D"ݿ_*FDD[(++^^^aaa4MVL`R͛7T*Upp0D.ƲX,@eeWqq{HHN+**tH$z!,&l h4\.:vhU,+44>px䉏ϭ[\nTTUrөS'?BPє(.Ι3zjС]v&);;tѢE.\(..NOOw333322ϟڵnذ`L2%,,F*8qfO6!,,l'NP*Vp8_}Ո#QTTԻwu֙Lcj Ry,̟?q*zҥ;wSӍ9D">}۶m ^xqAAAtt}>}[.;;W^"h}IIIܹH$:x5/d[u6o)ziii!!!^^^ hF X,"رc<oΝ[lӧύ7Fm[ "---::Hҽ{j41c0LCOGK,ϟ8qbjj^ _o&%%ݻw/<<믿q{ɞ999o;s秦T.]̞=ū=#m~aȑ?H$+W3go%Kp7C ٽ{d _n]FFƼy,KǎE"ŋ Cu:R|W9( rj]]]@@gLL̽{p>v B (w0؆ zlFL&Rǎ5kƍgX={B%oݺաCJ84~O>W\/,,ѣdzUTT㬬O>qU\\x@@aI&m޼Ϝ9s֭2 񆆆 Fc}^oF,cvʹ4 lCQT Êl6aIII=RSN?~|VVf3l͞=͛-7ۘ|>_.h4Z\\z~aСC Ÿq:~x\^]]m-22o߾ ca {)//>|O>ҥKaGyѤe˖¾SLfϞ=alnnn!!!766QFm߾}+VʧcH"v$۷oVVO?%$$>}a>>>_~=B5jԾ}l:t0k֬4ww1cƔhѢ7xh4}wBКkqJ[nׯ0̙3> 44/T( .lhhR -իW_~˖-틉w$f"Ç7|>L&}ѦM(JBBBRR5~ZmM?СC4ĉu:… ׬Y3fg?~FWW۷VZ{ۗݻ6 T*_~$dܹfom 5I'bM,wڵ޺ Xq={g}fo?9ٳgȐ!>>>VᴬY&??`0teN@Eh4uyݻN@+`G&a0C+.ӧ[sss-Kbb"aR4==}ذaV ڜ:ujС᜜44رccƌh۷ogggwd25Qٳ*jΝ...O?{ȑx߂t:]UUӧ?~ g՚ܻwoذaݻwoղCj"|Q^^^w_ pD|&IĢEf̘N>=~ \޿R 7o90 .4Ldڴiׯ ۽{_~ uVfff@@@}}~D"FRw}w̙\.~Ɂo~MJh4*׮]UUU>>>hI@-BsioMyW6/ ///p8/S[[O< ;w.y}-[^#֭[G8qs#0l˖-t:O?]hѷ~;s̉'Ѥ_rǏ bYr%t{&;u$  4h+W\re̘1/_>BJdffԼ|xzVԥK3sѣGCChm)VVVFFF0 xɟEZ...XYY 11 ?d "2qr7{7nd:tСʨ( ***_ϯ{]]]m$;V9jUk愻::"K~'hG&LHKK5kP(Dzxb:ѣk׮__8>L[&O>>Ph/ 9 8rn̜9s9„h!~Z Zy-[wW.^F8'h@ "z^:d1%i@hf@ / 2D3#4t/Zj@ "}BkPMz$@ NhM&Vuٰ.u Am0#!hDžAt?0 Oǩ bY~}tttllQDc0-ـZT*6mxҥK[-/>L&ʂ}8ggg3&--mԩpӧOufڲeK]]hi4 6ddd <?Pd2|-^غ8!=zҥ]͞oZ0K,r @W!l5T*?`F@}}}vvvVTT<{믿NNNk.]tI&m޼yȑ7nI8ݻwFqذayyy;vcǎD"ĉ?Ss26mOJX,@=G򉲲>lÆ ťG;w=Ię3g a,gΜIAf:q //Z9soׯ_ZjРAIII}Sƍӧï]vȑ,Zhر"6W^D7߿РACa0@f͚{o޼yA}G}m۶:@ddd@@#""NJ^رchh*l&6W^yjo`0,={wL6b>?k4m۶M:…,[l۷x<aqqq{Ν_x_Z-$mQ8 NeZ{=FӥKWVrd-T* 5@-?!޹~y3LHZ~SNeffB .\n_x?… f~2eʃjjj,X@ܞް~I*3}twʕ]X,FjRUU'tV\_N1 *jjuHH B>|x޼yBP(=ĉqqqs Ϟ=[XWPv~s;u1FQ.j4p'$I p"##?~񂃃-kdk}=L&BZPSSCF"(%`_ ؙV2o=Hc+?^zY8JP*A<7߬>uTcccxx7x<F} 7o^bb"B)**[nϞ=+W,--ݼyA`ZMhpw' +++F#JmR> MU]] crD&mTɓm֯_Ȑxh?\5Ilrt9tP?T*#i !277 Z!-"4 r\kP=f{J*Ơ^Ferƍ/H8zjTTaW^͍OMM]reϞ=1 YTTԡCӧ'''k4xfϚ5~!P(FVaaaD`ַo_'ŶT*'|2vX 2f̘'N0z eddk0 #"//sMbkaǏ?y_~9m4 `Z7Қ δ}<]O?tܸq/^oVllX,RBX,nUѬSUڵo>}СC777@NSSSN:}t1|p\X^^>o|cǎeeeF̙3A={633SVWVVZ ><11ƍ ㏭y\`ȑ#ѣX9vfpݺu[SB___~E"0 77(!܀ vjjj44{G!ls౗ו+W`$9}t^^V<<<`úXXXرcGx~ƌnnn=z`2999ׯ_ k3V^]TThF |ƍS*Z͜9s޽p&...//P*;6220iҤAad2]]])SdddLwy'886N:]rڵk W^L&Tӧ9agϞWnnnSLS9s&11>l9TjkDnl[FJH&Q f4_5QQQ֟:t3&ჂaNT9&ƃBئheС3Lp'۷mgۻfYVqH$(),D;v~DK,AM+fjkjj.ˍB w*h"{kAlomG " RWT= k-%DoozZuʳ aܹslVjZE"*+LfJ!QRBQ@dgg!h4Z688TV>1l BGc !-  .['C2kp!D 2 qWWW3SI?%ku }UjCWN'HF5UU* ٫A<;YiF)//71A!D8'y>zyy N0ssǞAAAK.͛M(_|s,X@" bƌC޽;.q˖-AAAaaa]t1bmOݻw ޫW&J8nk ڱcG{̘1/P۷oȑ#7oܾ}KfBL8Ѻv9@خPnQB@v11Vl6+={@Fnd,qC1 fX۵Lj`%X,uݩh*++atرNפW_M|xCCBӧl&"..\8|0t`v^zM:uذaG3fǏOMM;'OӧϬY222;vsfff'|ܹSPPfeΜ90a$8pÆ +W$b޽111SN!!!ǏWyy՜h4Nhc*:}t|?--WZ <x~~[x0;o"ٳΝe03g\rի9NAA#.\X`3g Ä vڕDV%b۶mp?|ӦM~֭[׬Y3p@ Ne˖hfÍ2w#sNXXl|vP(d2٭[?{k=f[h06h/^HPD"۷0 h4T*522r^^^s΅q=z'ի}˗l>CjHt9AX,Vl6>{ni i ٚrǎׯ[4-8FGԐ-Wo44 yCd -OKKHxn0X/9s.]:c x>77wѣ{= RRR~&L ENNΊ+JJJlvbb"fCS H.]`=..ڵk)))MJ&-[hllu4l0]0a Nd2_۾m`UKh j*$P. FxNNӱt_K{뭷lْlwFLMM]t)=zw^S[[ۤܿO>9~xppÇ9B 믿Κ5k{* RtqqyAnݺޥKgsz+Vi&N6euĉ>}ׯ_:ujM4 ON6 M`0FWW϶ sj`&>ΣNMV(M#R`0 ٜXtE֬YĉZ'%%aիW'&&bi0.^Hbh4AAA0NX6cƌٳg`p~V:u(0 5 *ojy:ZҥKE0 0aBAAw}g2p>x aT*`0 2Lm!`6dsժU)))111999!!!pR)Ld2dXRRRn޼߭[7>="x^rWƻKH[֔śӷtZR2 ___ S60KVi8ۆD6Q>--!BǣGԩ@ bRPHSTJ/^ypq-!U'M #f 뉄c2AM c/ٰajgl6߹s[nVMas8NeBpܸqaZ!1BAeyP&k9 ҧ&z=P(0D"еk+hgOTSN* g5xI[%h2F=iZ 4[B!NVP Scbb.]{7|ׯKT*m]@kԠdXI2 ۷!t`pWϮPlTUUݾ}oT-Bp?l0d}Y+QI$zYIjkkqww꟮## H V\io-Æ sD+ }GKHRQQa aw}l}ivESJl6b.WXhmP*ZT_t!c]1Zm]])!mA$/hFϚ5kÆ RC -d˖~8rbfBFq>jZR[%8 YWG. X, 騲KJNϥvפmߒN5:a„Doo9s?ugZzVh4\.N-F£t(h4n}uuSjF\^<:%p `\^-{$Q"}̀SM 6AFFFK'wԩa&A! (ŗ ZW7Kʢ V[Katr@I dfm AVt\X"Thdvip1BMt>hZ-\D7jJ6߇"555\.DK$b!$kʃ}"Jn}-v['n]+;{_m>M$-hҩffBX,Vjz7 12U/ DAN3,UPUK  f^7j_۝rWjB4)wS"V.]'[?j7!,&HO7`"S5/'\[]o#Nb $m-r G40B`M dΉ\.o$lWӫ4:-B`+%Q *3* ^ERDA( $HcGv CEcŶWYiyD<}zyyh4ZLl//ExdpROT&(B3cO>C,vyt :a.,w^e)^^]g=фnݺa֢5a\aӹ yyy08JLtrC&Q8 ^w#ȳ[7T32jo{m۶?vssSt*FYtBkziԩ :Nrooz'5?޵ka): 5n6@oڴu]7n(7~{ZY(28+u"gT*0`d/zѯ*hNeްaFޯOW:qɓ'NS(#G,**:wAذT>.l%ZJA`-be~A ,8rȚ5k222Hee~wbDNAtܑA:ww;۹+ˊ.Nr^E@=.U<rh@C[j Cӧ&O&MdwXJmгE-Bswߍ76/--DDDQ]m=[TFr:9`hѢ'N\_ҹ\˦s"ZiDh49rd̘1ѠБu7o$ml)aRe@-BsrcǶܣ/--R!!! 5xHi) IaXVHHH߿Kˋ"oR¼Q--1z+%-Ia5. XUC\Re@;#鴲垾l.++ J?ِt*`T:9j=a؄ n&:eK ɶ&-B@m-AR  jhSCڂ&//Ivנ٥^s观俗K /lJeQ6ֵFeLz  .jZZ=vXtE"@rC/%a uҪA~~nݺq- Au5o %-h$ 4qFy[ ֔k֬!-X,V*5ȍ=5ПlH:i0։0֫wwuD `۽ܐKb/_:iǵP@ڵ+0kld>v/mA%Ivנy\.6]+$LZg*..oJ:a0Aլ,ٜG*5j-6" CCC@ gjk ڵkW ..*Ψ5Y mAgmAN5kT,oٲe޽LyD!$ϭct*d1(2A2yWf̘q}>ڿۛ'i{p}Z)6O]]Ap(Dm-!Poݺ8z6BWjd<_{{:E}l6dAAٳg+DnFJ;( ,WK&hr+11֭[(RьLsεBB)ARp~T77J34w%gϞ-(( =qhLHHPTA(J8:::99?n z":ɓD #(( ~熤S$WK .ˬ,s]]ѯ^!{!h4/j:#h^$ úvT*=zaPjT6c&''GGGtz`04&Ν;>|ZNMM;;S(*v"anfsyyyXXLTT熤SY™2φ2d<~(딂z_WJuCrIAt PXXHPaQT jӌ1ZAD<,k?ۿ#i:BOѣGFNqPScD2̸Y@1φ uנqK[(=$_ [^=+m[b+&cf;q]O3JJ[/qi 4-55ַGU`cuV_qznH:i2u}wƆaXffKN&yRUKH6gBZ-BM` K. ??记TڽdM^^:dgꫯ r!k >i2g)))̔6X7W6Ph^M{ȑ$: 2{YWjk g2FN@-h*lo0Yz}~~@;ݻaXxx8NYF ԛP(ٳgM͍lX,p8pUtHk׮F~ Z7j???1+![d|쵮.s6zyh縕Ne1^ȑ# BS[[jT ` bn9i4e2p8 }s B!P &ިm i tYZVJ^wAڂHܽ{7\>Ѽ1j>FDUOH0fB9p@$5۽|?CKVU*BQBiibr)U:e[(H[c ѱc&'-u^(ۏfW^p6V;!t*FaОYMoKxxxPPPumu N5 ^3p4T*ZeE}WCCC9tD"?b70>{F{ aqڌf8SF҆-?Net:=99yǎ5 &BT*5LЭLM-Neh\AÇ>>< tJz@di?٘7(((yD4# ,>ьXN`RrED4 R;>sX~R$мj#&9{~.",t&J&cbb~駪;xԃ:w/)#[T^t >}<{><<|Ν/s{ (h/fph}zӶNe2hT#55B!UAƶeH$R@@@K?#Q DbJ&ڵkv Pb28"\dɳ'y<(yD}JKK1 0L$\.,TX77=#srz TcMMzغu1cBRU1thKN%(WE4hsϿGD3RQQі BCDv_P}"992//jiphTlʈDhI۷#;f͚  %zā7'tcz/*0 sj`&]b4op( Ĭf<-- DKިW%NDKt* 0oD"mܸQ ,_fm-a6>BϐK'sW{A Z~gUQl󮣢B$p -ᾩKMM Ljl$Zs&' 7tիg H.#T])=sy;^\[Q~d֭[Ν;DĶmG BaXZRH+f#RSSMPb@j<ܞuȑoFUՍ;wJd2ieCٜcFa{uoR[9 c# Gxٳg[#D[d„ ;4 tq#_,aaбWCBMzz? 忦u}#G4t:}ر-bVU41 [v+W?>kw>lhL%;͎_[ h3YZhB_Y{;C 7oBezѼ`01낂dAp+/-aʟ.~eQی# ҽ{MB"f`s.믿ưon&ݻwIIO?Hš*Qsp CX]]=eʔJ\. ܹs\\D Z*۬YP Bmh޹y>>>117nԺ<͇1ra2 9RΟB JKN޳v[2+WtORh^O8dh,8aN;lذk׮z>---33sԩV F;fk@;?{vZj$}Z,ǎf KQѤo E&a5f3QYeV1h/ð,Tʺudee9r6L$&R*8!$$$>|`꫙ϟxȑw&h,]biΙr999xS]붜P4/drڵk;xu1 %U?ߣV\٢IXNT@`_҃ĉ$l݊/ZL^|l^R]NoD;?tׯwoٷoٳ%B]l\OTLJ" N^ {>xŇC("&SsNzkP"*Ϗ L0BaFZG !1 j8!X,YYYvھ}Y6m4yiӦmٲު! vâsZ(@гgg+2J f:"Z_wOŖ,!Z2)a֑BL9Jemc:88xǎ;wtww GaV F2dH3q r0(YX,lf`[bq{uxo-@Rcsq}&)Un [Pߋvgg33;3pWWWkkk(ԩSv{xx+,,Lէ=m+ϟ4S r H{qm̭HgTU"z ¼<}}}==25f~++B G&7 Sj}>*B"xƍÇ9իvرcGTTgϞ]XX}!0 :Cda N_$=!LE555555 XBgֶE(p¼+iS j*Mn#^">}};SN\.[:TI$۪K!==N7#SIzb1`hg7J۵k_] ΫRK_/|H" Fa|vl˖$OؑDqz~;MEB/#Fz(-]hĉwahh pww߿^s/ΦSEE $䲲7"|<ɸ~0ɩQZ_zׯ_..^dfH"Lum|ĤmT;Bd*ڑJ w/)|!@] mmHݻw[h7 To6㙙VVVt:WcV3"0ld۵KVSQhmX\zm:XElٲQFyzz:::8PюTRLB }nPZ.!jfH-1h ,Ԏ`v\,FZt=U.oP!J655m󏲪}ŋgnX[ ovj)/_ $./xuC8u,^7sR"dA`kVΜ9SKK̙-XA\&VEmPcj۶m_>}!`0!Ih48ofgMv%52<884O8 76I ٳ+E!tvv.-EDc]öYCCc,Ν Jx"yEiWU*uL~ϛfYZZ"G 5jk*i݇S$Wf:]qUxYY׈!ebbRR^eӚ9sڑ#Gtu%sb~ UJ*}bwL&C,ӳ#f608s&96l7sID_RC̉6ѱjyҗᔖ0̢m+B;;ƺ@@]]O;U꘲YYcTU^^^ՈJg}ˑ2KC]k=?<<]6Y}Fի8_\YZZ~+**X,3aBF&] .\!tA heR1 =6MNUZ|gO(/G HNF oo@܉@X?k$}hmakccUQZn+JMMmѢE_#LTKvm6RY...EnnXJ neXRw*BTPF xBBJuq\Sj}%Ktf zX)#1o_LJ)|&:Ŋҗ(33RQD+0 [`L&;rH.;rXz"TI'Go_&)sssH<9t"*[/Scǎ,_r)..>n>C'IqGSwU9H5=*==йsLD֙ehƌc``p9]]TCWmM8W{"Tcɓ7}p#8x<`0X}o<]*]믙^K׷c}McLP6P([ɓ'٥v鹹AG?H۷oɐYS]^_hV2,% aU-ϡk6à" \]й2NPU .|&ImDUYL++!tկQi\ B&R`g-VxNu2w:bM-Yv0ҲPUNo޼dNN.׮v)Z4k@ <8555)tYSw1׳L"Vf*B !}拓{./ʜsdHdV] H:2J&R-555*jzI,<`ɓ'%'J>2T 6XWKhhpkGwH$,ڶmܷg<#3T}^6+WT^^NFVll%K2t@6h_MAAAlL *禤lx#Az~Cm6y25~jjBcs{Y١Z!CLXK$\.wΜ9nݺpBd9{S7A/+C^!,/ZiCT] (z½O c`Ѩ88r0ϯ::-! XZebK(bbb0-]='CyϟX[ΝR)0׳ r xyZ>KZby[RH5o-t%:rG[nرݻI| L_ :ҒP,`=+aǣD"sNƙyf ]A }YZ7mZ.y`<uϋS.Qk)W"p|*>ܽ@ԶgLMMݺ5Сyx+!:"hѢɏ;_6Iu,&䖧ꮭ޽{gىbx+T^qZF$B"^ݡpu{C!H?<#)/RÇm[ccٳ%fѽB0-S4ͨӧOhTF،HKJ& !؈lo.Ŗ- ̘qK1ْ%ӥsHGKX6!&p#^paӭ[hҤA}z.ݘ?$V9 t:ײ&Or#ݻwpΏ+*LfkDIFUS)XlF @ɓ'iixD/I w'ڰaCJJ6SgLoH$nBX(/!n@"R^F^r)kaaa7oNH–.LM*޼ysرF9, w6r>'|J.]޾} ,RIm2XG\- Xx54Y_e4 nC "lB54 CC1c,tK bep{SSay,o{wXgN#u Ͼ(~Flr)}^iKz8*0E:RWw,4p>n!Eq5ܲH2QءZ{ݲeKYY9`Xl4 C8p`ee<<aFFw.\uuAZiiiFJgΜ3f tzHHW^YH$"!deeٳ*ȩɤ O>H${Pm"ӫBܹs5T"7 [ھTΟPu*e𣢜%%%-*,, y z޼^n]C-immc:tgSbĴ705Ln\ʐ-gjΟ??**jƍW^q<66`g!N***ϟo 23[񼜗<@fd!lhh?~<޿s[cf3]-!9r$aaani[<0_Rw<`EZMZt}ui 8t; kmLOP{At$J҄;͛MMMk׮z]di`qUN2elo{F׹%ȥ𴴊EH !LtttEа}v B&ϟ?/᝝7H&kQ2:vL@"̝;7++k޼y3f̀=͉ׅNEo;rwSqMߙvƮ6lLv횋  KnE:ǏkhhxyyՁ&aFe ԩ &]U(2o6I'Ovq=~e HDq,BM4 B!=l썻(CB?~|ҤAMM;ʥ'Gjաun6gpW3e&M %SЕo~*o>qq? ؟6n܄t}iR<,88XPTu 8k׮={˷DDD ?"D"fSS<ɓ͛3!A6#P($L&GEE łqN=2\Rޓ[/OoCR.O,gϞca!!9lɁT_I{WiiӚ7oÇ:e޲f1ͽ)*J6y``7ԩS%uIҲ%LfUUUCoHѯ{/JᰭTPwj;E{vDDDDDD|3m"bjӗCv6bPU02RrQZQTRuà9 aT\àw: \Q׮V&H B$KB'NJ%cG?ihΜq"õKpə\LiiiA?>J.;o2""d5oJ %;;[e"۲Hٳ%gp7nqyoquζkol|OLre:jV|IlVIO d^^2uVkkAe. ̽W.LT(L1r2\_F|@ x?dsר=p&˖-ϣo!'O3-EG?86& k%&&7n͚5><ʳC=Logj ƍsrvʮDbs>(B ]] )>}mDGG!9fF4յ!TSyӃ0/etJxmGS8{k tx,, ! weXGeIU=m=]Q׮] zjz:zZ$''7СC%%%pe*vwZ2龍\ LMElat41qmDĥݻ=z}!t] ޮ# v;ursv4Ħ|`"[;\W*n޼944N)"e2m8QTVSʐJ5Zf ZnudDpoxqMީ$oر~TfaJ1I_[KkkO[{İ?8{)S/XȈdN<;eo~k`R_9!mlM2]0:Z0'g.]^`0ͫL-O9mMc*_OYL/.N3}F<xzr |Lʯ8cƌuefrqQ[e4 %鎎ܸϵ9>Ojd$+[$\oasy.ٳܹcooi޺ޔHT %aGc99ikT8y"%%eժU򲶶PSfv["sFO8x= ^Qp&8ʕ/+*{߿аXNE3:9]ǎ}0u4Xxrs=Z)JA!5kVs;}ܹsZZ3Ǎm΄ s=MAapmN Çc5kffW?}*ߘ9B28l_P(SLD+Wp 9oax ̙Y8N8hUW[j(C*>FF_?~]cǴL7oݾOxտezKlm*6$`4`J55/~qذi}###?I-"ԽW ׮%XHքݴYH$ky [EX)K{=QSN8@:tDlȑXX% -+L2kffOә"֋v8Wge,jɱ\.`g'MRPVUU <`5պv}׃Z^_|Uu}6!4҂pϞUTckR ֑(Ξs"AYwX1gEA37LL&6 ?c޷ؙ~xMM^mEQKve*6l]=8Xus'(yU$ZB^^ïjihoLX*쐃m7paXg;K_^f[ajj,bcьz (z5y~A`XC IMM]pAbb_~!Ca^E]̽x`#H$EBMc\R%HЛ7lGFF &lkRt\zR>rhw;e崰nHt>Jzh(}udB{{{ervy"n񅼼/Nn@P|/_ muu TWW8A<8G$Pݦ9vxrE99H.ZH<OݴiE'KnݒK6)gݕ%r:M!Q.>5"C~]k]<]X$ ٶI<9vB1_&PzׯeOʢdW^ 8np-N!4p}.CټWLsHҔG7oSQǓYV?5h-q0԰\\^E6n7G_62T*(*BPߔI&=ydȐ!G9}gYQO3{MbrصkAOn{fc ի8.۷74q١d䀀'뻴(C>.;l;;B#- =,k|;j]Èe&y-q u vmmxz}OeIˆ|>ْ%҃^S!ޥNFF /(1 ާO .<@sDޫи7&\jpߞ=/Hvz[p?׮L& ,11;00PM_3[]ͺ}*r)],,횋06t*M+Jeǭ :5TH* V_Þ=wx)5wEN6,4t uuXYA.>m ;T bc*bA2 rs\Vo#BBBN>Mя{6[ +'.3յk?EG߉ixqעduep̬we˨|~92?R,N\ݵ7R31NZ%g{L{o~: u H"ӦMc6vi]] a>9O (**_B"'Aʵ,;;p)CY9%{7~,^\Щ411 LN}hٲe/^Ѩ+pqM OieevCiwŗJC !C+wƍ6vM7Ǿ Xʥ NZh迅atfai: =0 X,VpM ͍u,+g۔$FQQ@x12y[a~뇩),ږb1ؽ/,DFƼ0hcŭ 5=<;6 q`v|xHprR2,+$J1ܔr?/]JHN7dȐ$46vq9̍}2Uv7j{o4Ե=H&*.I$_mmTȉa͛W^X/q8f%yrsS߾MP ³֕Vq$AiRoC h6mL01ͮL&uV׮]7nH"/_;xpmp>]nՙ%}fh e H8y;fvg?@)ܕѹs792jQ%coܺ:h}(LJaܔ^m\b)28i9;"##t0L|.V~ʷ<ϺhuWw!RRR~qY/qDNNGG9u霘 ^z7X怯yHLLZ[GxګWlaܹsB_%V10']:JWBJ8҃}&脭)-[$dWtaX ޛ'kY`gHI\-*T-B">sT`jbn:F 9ڵ~A4Ttlxׯ<<`tVXcYYB555 .twwp|B655;ǎ-2Dd2T`oo7n=jjj A k낧;Wx*ff}H9#~|UK&l6ڵK24tI\ ::{S7sz*}* R'-g5Ʈzv)FyrTCD"q }Nx">>>##w^XXT*;vlzz7v.e͛`S|Q})Z{U֯_!DuF2 7ήXt*ZVWfUg;{9W話O(t h(a _diRt!q>fKKYeeUNmcccSSScc#p8,K,hck!W,x70:2-߶ҥd˗O~&5!H4o523IF)Iw+}G2`nR/+(4VV**:֭ɓjh)ѣS VNLMiWV#\`Att4jq=J$#F?Ǣ`FuUo8 gbx)uy;))lE(i^@1 7;v={v"" ^[Qm%ct]_DOi4ڒWG߭ǣ'T}BΨ%X[ gdttF4s٘>2ad-Vj8Q~`n ̘ak``gee,aXYYaѣ:IIn܈d05܊*Ce\$*Cm܆7nܢE\(&3oNi]|IjkM!ɡ÷l&@*K*-j-ץ3rҪg,=P,ǰ =u+Rϟkjzfmɔ"H ݻ7o޼;w2M~?~^OCY9GM @N8qٲ2.Oמ.\P''O?4T&E}jm؉IQ#) VqIm>(D-[V^me>Ǣ OvS5^Rd2t7oOXg_7To M0N;v̙D"qf D PYYŋ^z% İ.]ܾn``zmmयi矸&8Bz,o! l$ZZZtuq*mp첳utt+ܕ2P$M06FaÅ =יRx)(9o߾ 1tb>6kppLnjSQQtQ+++U]RNŌ˯x/C9<*Yy{wUru%KbccR)Htrr0aP[JV[M斧v2u:Imɱ܊ӷCs҉HFåF$AWWk;t=Z;ڵ<OKKe\p%O}q @%pC ?*"p\&x6n6MB1_(拥"B8yY1L4nJKR) 7lid Ǯ! x$>xسieLMf||g^|ٳggg*XYsDvիWF_dpB BNEoa>DT@]Ut)t !RTu UUgG$4#du*agL6ELZa۱g)((زeٳg---_ŗq?"h:|gCo6GΜ9S+(>j#N59U{rR ҊsSM_W˨>"@"$-XOph?C]uǒJdž  5k-[`#TkA]ݲ 8B IxBKO~:-#D*B฀^1X8. ނ#YrS.'$th>qĪU444_w7͛P0|rNp%eUl+v++NL\]V`4 p(lh6E׉_GuRt (ٰtqǦ|wtWTBngH$T*y<^IIYQ~'dulh:&g@ B@SSBH$@*]daa99݋N".g/)sg8d>JW>*&M҅ak~lۿwHSRtO qzƊR(492ܴ 艈ά1xǎq/I#7dop4gea...>>>ii?0c2!B:mmn'Zt !d `a՛&\leLT;,$ݝ MFR0d`&! qd1 < fPjQ>xN.N}A$2Δnvuha3Y|Dl0_iJ[#m }{_>jesdaTN,!);i ׯWDMEQE2ӵ2pT X5c A @iRtm!Uc@mOm^ne];T ۷("EF˱ W~ѶF.#ukeg14o']0ձҢn8 ;y>>C[NGס{նC9::bp'*a 5jp[dlF#\ %ǣDBj=Ɍ*sͽda^u6*MzUtb:H$wQUANZEըQ;&>|8tP ~TB:ݫfYlgLMCa|!mĞK,z?N*vɅ'{y| ӉX,~|θ2R|nr)Ug!^SuFpFP,];W1-' O0H;W]]_v3rx4`~\.^O3o'}K Tj+IҸ8EP-U *)J"vR[JUJ*]r)UTcP.epll9C| gGF>ڵkORJ]]}ʕKjwCTT<*#]Mw4K"(V"E "b2:DI&稃ѸQј#c0Dd`@ 4 on^GuL~r֭VgZ܄L&ʹ #D o9oAAA{3b&r'-F[V! s!LNNf)))&&&f5B.\@x/r@H+W>leeс⅃ޱs! bggPJrCL7FRo/RhS剨j kV^=\.P(Dd233k׮?Y3G}4* &i-f#ʹ?2 Z-_@d2׬Yf͚U @ ?xuxjzppАLߥ====dh{ L͵rrrjkkKNN޿!t0RT{ T+rtq|ddL&ϛ7ozӍrohhdNuvv[ZZL&p8?Յ o```b<%:::TtiI.bq\ tvvjiiMڝ$Ȉ ±1r߿/MMM'>@޽;::jffFJewwTV(MMM<266RӕE&wuu*&ꪫE")VT]]]<000 SPvSSӉlkk0lhh`LJEd2||vxb* W8yǎg777[[ۍ7͛7믿r ߿rʄ+V:t_\\pBss;wN} %c||VRAD"Y~֭[###}}}% wwwY>::߶/ C<<ƍ|7n8pڵkX,VttRqcҥqqq|۳g\.[h xAAƍ{=+υ\.x7r9gΜ ݽ{U""" \\\k׮(JU*hhhttի766P(/^L')a&9222 ȈD"$BPA`}I(r===uuuDC( MLL`lddRt2L"d ؘRA hBaCCĆ0--- }```ZZ~2L,CפH:22=!r~~~TP!"sK6bZm6!0ɩEIR(gX8.,XP]]Ν;?aؘuKK mǎEEEІ=z4!v/|||& Fh^OYP/tqΞ۷Looo<< 8. j̙3>B!H$ qb,///$$7tuuNQGGNL3 *m77  foذ_yy9̟?aN-ܐ͉pӃxs><<_{5T*4551LSSSbڤ6a~~~$`\fRiuumϞ0 d2[VV&\[[`0LLL8u5E$544xx ~~~0@^^{֤޽{: ΗvT-,,|!N _P(gϞݼyXvv6䄆>1ȳof`3Lxmll<<>xy*:1g`<~͛7]qqƍ濷W&_Nq[XXLֈͤWBfxwv@WWWXXǃmQQ3000GEDDV𷀵\.xX<::jeeEP>}(33.,P(/aurrqׯƽH$BY`OmmyT5LRNNNK.mnnh4!!!D@ `0kjjD<ͰUu*###7o۷oyIɜloo߾-+**ݻhhhزeKll,/))iiioڴ +))MMMrʊD" ]zL&/\%H$Ǐ KJJ~ǖˆ"ߗ奥REDDdeeUVVVVVfggoݺU"xyy|~PP^iiiUUlBqΝ@AA$kx=266i&ssP ^?#\.Vm*~ԩO>w)|˗/9s&>>dxxD"ƞ={ƍBǧ){Anjj|}FFƳ'***33p8h@FFFuuu[[۵kWZ>|xUmmmAAA{{O>a@DDĿw\~}۶m8[L&ðp8{as8M6^ztt4((nה´~7nnnFFFiii666 %77.+LLLMMM;;;-,,D"QzzׄIffP(o/_z1@jje޽GR\.~:`ҥ, ;00|-$kYd a-j4x=<((jIӣ}].\xJ^SS3qNܹs' fs܉4ޚ:R!ͪbeܹH$G4A^^y \.gY 7aG9re󐒒uV8:/-ZA1 yZBC@ /f"RsLT*>onn9g, ###_O! `F@ڭi9_yfx._|J---'.P(uuuhb1 8;;kiiԴjkkbѣG:::˖-# /Wigb:h+Vwާ~W_޽lŊ C"[L&:33g{{%K.^c#H\.766ѣD;rHEE{ggkkkE"ӧT_~`0ܹsĉǏ.[liiivvv0NJFјLfoo.DO훛,Xp6scccǎD~bG# @<ؔ^oooHki4?III%''ٳt޽{rs}}=aJb566¡Ѳ2wwI{pe25qajZ$ݽ{ðF>~Iqfdd30 ;p@HH$JGGG1 [nƍsrruxxx^^ޯ<㓗ggggnnrl;Nljj B}}}>ݝ eoooyggIx<<<#""97oLII144TPhcc3O [__vZ8*L";>>N-[FJΝ/?nii ~9@F!T֭[w!+++???}||y<q\WW7..6L& B:SԞKK˨/ׯ_g0o?wp8Gv?-aMHH{yKx INb(éS? F]t )p@V SfX$, 3e˖CCC:::\.399D"QZE-LNLL\z֭[R)F)++[jP(R F 8hjs0 ##b022fff H$ h4Q,--B&Y__/ÓT*FkkkxӔL^BRwfaaʕ+WZ kMPPT*J ...W^_vZtҥy222:{Çդn,k޽$)%%>H$~ixx8N_IIIYYY.\Xd Xaaaaaa<gee^722qѣIIIpaɓ'0 9wz8 ~#-_ ?K7qy?ϊZ=M s 2ә_K"J`j ?_ ~0 @V' endstream endobj 395 0 obj << /Type /Page /Parent 3 0 R /Annots [ 397 0 R 399 0 R 401 0 R 403 0 R 405 0 R 407 0 R 409 0 R 411 0 R 413 0 R 415 0 R 417 0 R 419 0 R 421 0 R 423 0 R 425 0 R 427 0 R 429 0 R 431 0 R 433 0 R 435 0 R 437 0 R ] /Contents 396 0 R >> endobj 396 0 obj << /Length 24691 >> stream 0.271 0.267 0.267 rg q 15.000 51.387 577.500 725.613 re W n 0.271 0.267 0.267 rg BT 26.250 767.476 Td /F1 9.8 Tf [(Our paper is the first to demonstrate that Twitter data improves influenza forecasts over what can be extracted from non-)] TJ ET BT 26.250 755.571 Td /F1 9.8 Tf [(retrospective ILINet data. Surprisingly, our study found that GFT hurt, rather than helped, forecasts. Previous studies have )] TJ ET BT 26.250 743.667 Td /F1 9.8 Tf [(found that GFT provides better surveillance results when compared to retrospective historic data. )] TJ ET 0.267 0.267 0.267 rg BT 446.758 745.174 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 456.395 747.555 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 458.804 745.174 Td /F4 8.7 Tf [(22)] TJ ET 0.271 0.267 0.267 rg BT 468.442 743.667 Td /F1 9.8 Tf [( One possible )] TJ ET BT 26.250 731.762 Td /F1 9.8 Tf [(explanation for this discrepancy is that our study was restricted to only three seasons, during one of which GFT performed )] TJ ET BT 26.250 719.857 Td /F1 9.8 Tf [(worse than usual; however, models trained on additional years of GFT data are not comparable to the Twitter-based models, )] TJ ET BT 26.250 707.952 Td /F1 9.8 Tf [(which were the focus of this study. Other reasons may be that previous studies used revised CDC data or assumed a 2-week )] TJ ET BT 26.250 696.048 Td /F1 9.8 Tf [(lag \(instead of the more accurate 1-week lag\). As we have argued elsewhere, )] TJ ET 0.267 0.267 0.267 rg BT 363.317 697.555 Td /F4 8.7 Tf [(26)] TJ ET 0.271 0.267 0.267 rg BT 372.955 696.048 Td /F1 9.8 Tf [( there are several benefits to using Twitter over )] TJ ET BT 26.250 684.143 Td /F1 9.8 Tf [(GFT, including the ubiquity, openness, public availability, and ease of use of Twitter data. These factors have led the wider )] TJ ET BT 26.250 672.238 Td /F1 9.8 Tf [(academic community to focus on Twitter, especially in light of recent poor performance of GFT, and the attendant concerns )] TJ ET BT 26.250 660.333 Td /F1 9.8 Tf [(about using metrics based on proprietary data and algorithms. )] TJ ET 0.267 0.267 0.267 rg BT 296.666 661.841 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 306.304 660.333 Td /F1 9.8 Tf [( As we collect additional years of tweets, we will be able to )] TJ ET BT 26.250 648.429 Td /F1 9.8 Tf [(make broader claims about the relative utility of Google and Twitter data. Furthermore, our results do not preclude new and )] TJ ET BT 26.250 636.524 Td /F1 9.8 Tf [(more sophisticated methods that rely on Google )] TJ ET 0.267 0.267 0.267 rg BT 235.972 638.031 Td /F4 8.7 Tf [(20)] TJ ET 0.271 0.267 0.267 rg BT 245.610 636.524 Td /F1 9.8 Tf [( or Twitter data.)] TJ ET BT 26.250 617.119 Td /F1 9.8 Tf [(While our experiments focused on national influenza prevalence, forecasting systems have much more utility at finer geographic )] TJ ET BT 26.250 605.214 Td /F1 9.8 Tf [(scales. Recent work has demonstrated that Twitter data correlate with ILI rates at the municipal level )] TJ ET 0.267 0.267 0.267 rg BT 461.929 606.722 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 471.566 609.103 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 473.975 606.722 Td /F4 8.7 Tf [(24)] TJ ET 0.271 0.267 0.267 rg BT 483.613 605.214 Td /F1 9.8 Tf [( , suggesting that )] TJ ET BT 26.250 593.310 Td /F1 9.8 Tf [(Web data could improve forecasts for cities as well. More sophisticated models are typically used in practice \(18, 25\), and our )] TJ ET BT 26.250 581.405 Td /F1 9.8 Tf [(encouraging preliminary results motivate the need for experimenting with Twitter and GFT data in richer models, such as those )] TJ ET BT 26.250 569.500 Td /F1 9.8 Tf [(that take full advantage of variables unique to social media \(e.g., daily, rather than weekly, ILI estimates\).)] TJ ET BT 26.250 532.898 Td /F4 12.0 Tf [(Competing Interests)] TJ ET BT 26.250 512.943 Td /F1 9.8 Tf [(Dr. Dredze reports receipt of compensation for travel for talks at various academic, corporate, and governmental entities and )] TJ ET BT 26.250 501.039 Td /F1 9.8 Tf [(consulting for Directing Medicine, Progeny Systems, and Sickweather. Mr. Paul serves on the advisory board for Sickweather.)] TJ ET BT 26.250 464.436 Td /F4 12.0 Tf [(References)] TJ ET BT 26.250 436.982 Td /F1 9.8 Tf [(1.)] TJ ET BT 38.132 436.982 Td /F1 9.8 Tf [(Chretien JP, George D, Shaman J, Chitale RA, McKenzie FE. Influenza forecasting in human populations: a scoping review. )] TJ ET BT 26.250 425.077 Td /F1 9.8 Tf [(PLoS One. 2014;9\(4\):e94130. PubMed PMID:24714027.)] TJ ET BT 26.250 405.672 Td /F1 9.8 Tf [(2.)] TJ ET BT 38.132 405.672 Td /F1 9.8 Tf [(Nsoesie E, Mararthe M, Brownstein J. Forecasting peaks of seasonal influenza epidemics. PLoS Curr. 2013 Jun 21;5. )] TJ ET BT 26.250 393.768 Td /F1 9.8 Tf [(PubMed PMID:23873050.)] TJ ET BT 26.250 374.363 Td /F1 9.8 Tf [(3.)] TJ ET BT 38.132 374.363 Td /F1 9.8 Tf [(Shaman J, Karspeck A, Yang W, Tamerius J, Lipsitch M. Real-time influenza forecasts during the 2012-2013 season. Nat )] TJ ET BT 26.250 362.458 Td /F1 9.8 Tf [(Commun. 2013;4:2837. PubMed PMID:24302074.)] TJ ET BT 26.250 343.053 Td /F1 9.8 Tf [(4.)] TJ ET BT 38.132 343.053 Td /F1 9.8 Tf [(Soebiyanto RP, Adimi F, Kiang RK. Modeling and predicting seasonal influenza transmission in warm regions using )] TJ ET BT 26.250 331.149 Td /F1 9.8 Tf [(climatological parameters. PLoS One. 2010 Mar 1;5\(3\):e9450. PubMed PMID:20209164.)] TJ ET BT 26.250 311.744 Td /F1 9.8 Tf [(5.)] TJ ET BT 38.132 311.744 Td /F1 9.8 Tf [(Culotta, A. Towards detecting influenza epidemics by analyzing Twitter messages. In ACM Workshop on Social Media )] TJ ET BT 26.250 299.839 Td /F1 9.8 Tf [(Analytics. 2010.)] TJ ET BT 26.250 280.434 Td /F1 9.8 Tf [(6.)] TJ ET BT 38.132 280.434 Td /F1 9.8 Tf [(Paul, MJ, Dredze, M. You Are What You Tweet: Analyzing Twitter for Public Health. In International Conference on Weblogs )] TJ ET BT 26.250 268.530 Td /F1 9.8 Tf [(and Social Media \(ICWSM\). 2011.)] TJ ET BT 26.250 249.125 Td /F1 9.8 Tf [(7.)] TJ ET BT 38.132 249.125 Td /F1 9.8 Tf [(Lampos V, Cristianini N. Nowcasting Events from the Social Web with Statistical Learning. ACM Transactions on Intelligent )] TJ ET BT 26.250 237.220 Td /F1 9.8 Tf [(Systems and Technology; 2012 Sep 1;3\(4\):122. DOI: 10.1145/2337542.2337557)] TJ ET BT 26.250 217.815 Td /F1 9.8 Tf [(8.)] TJ ET BT 38.132 217.815 Td /F1 9.8 Tf [(Dredze, M. How Social Media Will Change Public Health. IEEE Intelligent Systems; vol. 27, no. 4, pp. 8184, Jul. 2012.)] TJ ET BT 26.250 198.411 Td /F1 9.8 Tf [(9.)] TJ ET BT 38.132 198.411 Td /F1 9.8 Tf [(Chew C, Eysenbach G. Pandemics in the age of Twitter: content analysis of Tweets during the 2009 H1N1 outbreak. PLoS )] TJ ET BT 26.250 186.506 Td /F1 9.8 Tf [(One. 2010 Nov 29;5\(11\):e14118. PubMed PMID:21124761.)] TJ ET BT 26.250 167.101 Td /F1 9.8 Tf [(10.)] TJ ET BT 43.553 167.101 Td /F1 9.8 Tf [(Salath M, Khandelwal S. Assessing vaccination sentiments with online social media: implications for infectious disease )] TJ ET BT 26.250 155.196 Td /F1 9.8 Tf [(dynamics and control. PLoS Comput Biol. 2011 Oct;7\(10\):e1002199. PubMed PMID:22022249.)] TJ ET BT 26.250 135.792 Td /F1 9.8 Tf [(11.)] TJ ET BT 43.553 135.792 Td /F1 9.8 Tf [(Lamb, A, Paul, MJ, Dredze, M. Separating Fact from Fear: Tracking Flu Infections on Twitter. In North American Chapter of )] TJ ET BT 26.250 123.887 Td /F1 9.8 Tf [(the Association for Computational Linguistics \(NAACL\). 2013.)] TJ ET BT 26.250 104.482 Td /F1 9.8 Tf [(12.)] TJ ET BT 43.553 104.482 Td /F1 9.8 Tf [(Gesualdo F, Stilo G, Agricola E, Gonfiantini MV, Pandolfi E, Velardi P, Tozzi AE. Influenza-like illness surveillance on )] TJ ET BT 26.250 92.577 Td /F1 9.8 Tf [(Twitter through automated learning of nave language. PLoS One. 2013;8\(12\):e82489. PubMed PMID:24324799.)] TJ ET BT 26.250 73.173 Td /F1 9.8 Tf [(13.)] TJ ET BT 43.553 73.173 Td /F1 9.8 Tf [(Ginsberg J, Mohebbi MH, Patel RS, Brammer L, Smolinski MS, Brilliant L. Detecting influenza epidemics using search )] TJ ET BT 26.250 61.268 Td /F1 9.8 Tf [(engine query data. Nature. 2009 Feb 19;457\(7232\):1012-4. PubMed PMID:19020500.)] TJ ET Q q 15.000 51.387 577.500 725.613 re W n 0.271 0.267 0.267 rg BT 26.250 767.476 Td /F1 9.8 Tf [(Our paper is the first to demonstrate that Twitter data improves influenza forecasts over what can be extracted from non-)] TJ ET BT 26.250 755.571 Td /F1 9.8 Tf [(retrospective ILINet data. Surprisingly, our study found that GFT hurt, rather than helped, forecasts. Previous studies have )] TJ ET BT 26.250 743.667 Td /F1 9.8 Tf [(found that GFT provides better surveillance results when compared to retrospective historic data. )] TJ ET 0.267 0.267 0.267 rg BT 446.758 745.174 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 456.395 747.555 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 458.804 745.174 Td /F4 8.7 Tf [(22)] TJ ET 0.271 0.267 0.267 rg BT 468.442 743.667 Td /F1 9.8 Tf [( One possible )] TJ ET BT 26.250 731.762 Td /F1 9.8 Tf [(explanation for this discrepancy is that our study was restricted to only three seasons, during one of which GFT performed )] TJ ET BT 26.250 719.857 Td /F1 9.8 Tf [(worse than usual; however, models trained on additional years of GFT data are not comparable to the Twitter-based models, )] TJ ET BT 26.250 707.952 Td /F1 9.8 Tf [(which were the focus of this study. Other reasons may be that previous studies used revised CDC data or assumed a 2-week )] TJ ET BT 26.250 696.048 Td /F1 9.8 Tf [(lag \(instead of the more accurate 1-week lag\). As we have argued elsewhere, )] TJ ET 0.267 0.267 0.267 rg BT 363.317 697.555 Td /F4 8.7 Tf [(26)] TJ ET 0.271 0.267 0.267 rg BT 372.955 696.048 Td /F1 9.8 Tf [( there are several benefits to using Twitter over )] TJ ET BT 26.250 684.143 Td /F1 9.8 Tf [(GFT, including the ubiquity, openness, public availability, and ease of use of Twitter data. These factors have led the wider )] TJ ET BT 26.250 672.238 Td /F1 9.8 Tf [(academic community to focus on Twitter, especially in light of recent poor performance of GFT, and the attendant concerns )] TJ ET BT 26.250 660.333 Td /F1 9.8 Tf [(about using metrics based on proprietary data and algorithms. )] TJ ET 0.267 0.267 0.267 rg BT 296.666 661.841 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 306.304 660.333 Td /F1 9.8 Tf [( As we collect additional years of tweets, we will be able to )] TJ ET BT 26.250 648.429 Td /F1 9.8 Tf [(make broader claims about the relative utility of Google and Twitter data. Furthermore, our results do not preclude new and )] TJ ET BT 26.250 636.524 Td /F1 9.8 Tf [(more sophisticated methods that rely on Google )] TJ ET 0.267 0.267 0.267 rg BT 235.972 638.031 Td /F4 8.7 Tf [(20)] TJ ET 0.271 0.267 0.267 rg BT 245.610 636.524 Td /F1 9.8 Tf [( or Twitter data.)] TJ ET BT 26.250 617.119 Td /F1 9.8 Tf [(While our experiments focused on national influenza prevalence, forecasting systems have much more utility at finer geographic )] TJ ET BT 26.250 605.214 Td /F1 9.8 Tf [(scales. Recent work has demonstrated that Twitter data correlate with ILI rates at the municipal level )] TJ ET 0.267 0.267 0.267 rg BT 461.929 606.722 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 471.566 609.103 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 473.975 606.722 Td /F4 8.7 Tf [(24)] TJ ET 0.271 0.267 0.267 rg BT 483.613 605.214 Td /F1 9.8 Tf [( , suggesting that )] TJ ET BT 26.250 593.310 Td /F1 9.8 Tf [(Web data could improve forecasts for cities as well. More sophisticated models are typically used in practice \(18, 25\), and our )] TJ ET BT 26.250 581.405 Td /F1 9.8 Tf [(encouraging preliminary results motivate the need for experimenting with Twitter and GFT data in richer models, such as those )] TJ ET BT 26.250 569.500 Td /F1 9.8 Tf [(that take full advantage of variables unique to social media \(e.g., daily, rather than weekly, ILI estimates\).)] TJ ET BT 26.250 532.898 Td /F4 12.0 Tf [(Competing Interests)] TJ ET BT 26.250 512.943 Td /F1 9.8 Tf [(Dr. Dredze reports receipt of compensation for travel for talks at various academic, corporate, and governmental entities and )] TJ ET BT 26.250 501.039 Td /F1 9.8 Tf [(consulting for Directing Medicine, Progeny Systems, and Sickweather. Mr. Paul serves on the advisory board for Sickweather.)] TJ ET BT 26.250 464.436 Td /F4 12.0 Tf [(References)] TJ ET BT 26.250 436.982 Td /F1 9.8 Tf [(1.)] TJ ET BT 38.132 436.982 Td /F1 9.8 Tf [(Chretien JP, George D, Shaman J, Chitale RA, McKenzie FE. Influenza forecasting in human populations: a scoping review. )] TJ ET BT 26.250 425.077 Td /F1 9.8 Tf [(PLoS One. 2014;9\(4\):e94130. PubMed PMID:24714027.)] TJ ET BT 26.250 405.672 Td /F1 9.8 Tf [(2.)] TJ ET BT 38.132 405.672 Td /F1 9.8 Tf [(Nsoesie E, Mararthe M, Brownstein J. Forecasting peaks of seasonal influenza epidemics. PLoS Curr. 2013 Jun 21;5. )] TJ ET BT 26.250 393.768 Td /F1 9.8 Tf [(PubMed PMID:23873050.)] TJ ET BT 26.250 374.363 Td /F1 9.8 Tf [(3.)] TJ ET BT 38.132 374.363 Td /F1 9.8 Tf [(Shaman J, Karspeck A, Yang W, Tamerius J, Lipsitch M. Real-time influenza forecasts during the 2012-2013 season. Nat )] TJ ET BT 26.250 362.458 Td /F1 9.8 Tf [(Commun. 2013;4:2837. PubMed PMID:24302074.)] TJ ET BT 26.250 343.053 Td /F1 9.8 Tf [(4.)] TJ ET BT 38.132 343.053 Td /F1 9.8 Tf [(Soebiyanto RP, Adimi F, Kiang RK. Modeling and predicting seasonal influenza transmission in warm regions using )] TJ ET BT 26.250 331.149 Td /F1 9.8 Tf [(climatological parameters. PLoS One. 2010 Mar 1;5\(3\):e9450. PubMed PMID:20209164.)] TJ ET BT 26.250 311.744 Td /F1 9.8 Tf [(5.)] TJ ET BT 38.132 311.744 Td /F1 9.8 Tf [(Culotta, A. Towards detecting influenza epidemics by analyzing Twitter messages. In ACM Workshop on Social Media )] TJ ET BT 26.250 299.839 Td /F1 9.8 Tf [(Analytics. 2010.)] TJ ET BT 26.250 280.434 Td /F1 9.8 Tf [(6.)] TJ ET BT 38.132 280.434 Td /F1 9.8 Tf [(Paul, MJ, Dredze, M. You Are What You Tweet: Analyzing Twitter for Public Health. In International Conference on Weblogs )] TJ ET BT 26.250 268.530 Td /F1 9.8 Tf [(and Social Media \(ICWSM\). 2011.)] TJ ET BT 26.250 249.125 Td /F1 9.8 Tf [(7.)] TJ ET BT 38.132 249.125 Td /F1 9.8 Tf [(Lampos V, Cristianini N. Nowcasting Events from the Social Web with Statistical Learning. ACM Transactions on Intelligent )] TJ ET BT 26.250 237.220 Td /F1 9.8 Tf [(Systems and Technology; 2012 Sep 1;3\(4\):122. DOI: 10.1145/2337542.2337557)] TJ ET BT 26.250 217.815 Td /F1 9.8 Tf [(8.)] TJ ET BT 38.132 217.815 Td /F1 9.8 Tf [(Dredze, M. How Social Media Will Change Public Health. IEEE Intelligent Systems; vol. 27, no. 4, pp. 8184, Jul. 2012.)] TJ ET BT 26.250 198.411 Td /F1 9.8 Tf [(9.)] TJ ET BT 38.132 198.411 Td /F1 9.8 Tf [(Chew C, Eysenbach G. Pandemics in the age of Twitter: content analysis of Tweets during the 2009 H1N1 outbreak. PLoS )] TJ ET BT 26.250 186.506 Td /F1 9.8 Tf [(One. 2010 Nov 29;5\(11\):e14118. PubMed PMID:21124761.)] TJ ET BT 26.250 167.101 Td /F1 9.8 Tf [(10.)] TJ ET BT 43.553 167.101 Td /F1 9.8 Tf [(Salath M, Khandelwal S. Assessing vaccination sentiments with online social media: implications for infectious disease )] TJ ET BT 26.250 155.196 Td /F1 9.8 Tf [(dynamics and control. PLoS Comput Biol. 2011 Oct;7\(10\):e1002199. PubMed PMID:22022249.)] TJ ET BT 26.250 135.792 Td /F1 9.8 Tf [(11.)] TJ ET BT 43.553 135.792 Td /F1 9.8 Tf [(Lamb, A, Paul, MJ, Dredze, M. Separating Fact from Fear: Tracking Flu Infections on Twitter. In North American Chapter of )] TJ ET BT 26.250 123.887 Td /F1 9.8 Tf [(the Association for Computational Linguistics \(NAACL\). 2013.)] TJ ET BT 26.250 104.482 Td /F1 9.8 Tf [(12.)] TJ ET BT 43.553 104.482 Td /F1 9.8 Tf [(Gesualdo F, Stilo G, Agricola E, Gonfiantini MV, Pandolfi E, Velardi P, Tozzi AE. Influenza-like illness surveillance on )] TJ ET BT 26.250 92.577 Td /F1 9.8 Tf [(Twitter through automated learning of nave language. PLoS One. 2013;8\(12\):e82489. PubMed PMID:24324799.)] TJ ET BT 26.250 73.173 Td /F1 9.8 Tf [(13.)] TJ ET BT 43.553 73.173 Td /F1 9.8 Tf [(Ginsberg J, Mohebbi MH, Patel RS, Brammer L, Smolinski MS, Brilliant L. Detecting influenza epidemics using search )] TJ ET BT 26.250 61.268 Td /F1 9.8 Tf [(engine query data. Nature. 2009 Feb 19;457\(7232\):1012-4. PubMed PMID:19020500.)] TJ ET Q q 15.000 51.387 577.500 725.613 re W n 0.271 0.267 0.267 rg BT 26.250 767.476 Td /F1 9.8 Tf [(Our paper is the first to demonstrate that Twitter data improves influenza forecasts over what can be extracted from non-)] TJ ET BT 26.250 755.571 Td /F1 9.8 Tf [(retrospective ILINet data. Surprisingly, our study found that GFT hurt, rather than helped, forecasts. Previous studies have )] TJ ET BT 26.250 743.667 Td /F1 9.8 Tf [(found that GFT provides better surveillance results when compared to retrospective historic data. )] TJ ET 0.267 0.267 0.267 rg BT 446.758 745.174 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 456.395 747.555 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 458.804 745.174 Td /F4 8.7 Tf [(22)] TJ ET 0.271 0.267 0.267 rg BT 468.442 743.667 Td /F1 9.8 Tf [( One possible )] TJ ET BT 26.250 731.762 Td /F1 9.8 Tf [(explanation for this discrepancy is that our study was restricted to only three seasons, during one of which GFT performed )] TJ ET BT 26.250 719.857 Td /F1 9.8 Tf [(worse than usual; however, models trained on additional years of GFT data are not comparable to the Twitter-based models, )] TJ ET BT 26.250 707.952 Td /F1 9.8 Tf [(which were the focus of this study. Other reasons may be that previous studies used revised CDC data or assumed a 2-week )] TJ ET BT 26.250 696.048 Td /F1 9.8 Tf [(lag \(instead of the more accurate 1-week lag\). As we have argued elsewhere, )] TJ ET 0.267 0.267 0.267 rg BT 363.317 697.555 Td /F4 8.7 Tf [(26)] TJ ET 0.271 0.267 0.267 rg BT 372.955 696.048 Td /F1 9.8 Tf [( there are several benefits to using Twitter over )] TJ ET BT 26.250 684.143 Td /F1 9.8 Tf [(GFT, including the ubiquity, openness, public availability, and ease of use of Twitter data. These factors have led the wider )] TJ ET BT 26.250 672.238 Td /F1 9.8 Tf [(academic community to focus on Twitter, especially in light of recent poor performance of GFT, and the attendant concerns )] TJ ET BT 26.250 660.333 Td /F1 9.8 Tf [(about using metrics based on proprietary data and algorithms. )] TJ ET 0.267 0.267 0.267 rg BT 296.666 661.841 Td /F4 8.7 Tf [(17)] TJ ET 0.271 0.267 0.267 rg BT 306.304 660.333 Td /F1 9.8 Tf [( As we collect additional years of tweets, we will be able to )] TJ ET BT 26.250 648.429 Td /F1 9.8 Tf [(make broader claims about the relative utility of Google and Twitter data. Furthermore, our results do not preclude new and )] TJ ET BT 26.250 636.524 Td /F1 9.8 Tf [(more sophisticated methods that rely on Google )] TJ ET 0.267 0.267 0.267 rg BT 235.972 638.031 Td /F4 8.7 Tf [(20)] TJ ET 0.271 0.267 0.267 rg BT 245.610 636.524 Td /F1 9.8 Tf [( or Twitter data.)] TJ ET BT 26.250 617.119 Td /F1 9.8 Tf [(While our experiments focused on national influenza prevalence, forecasting systems have much more utility at finer geographic )] TJ ET BT 26.250 605.214 Td /F1 9.8 Tf [(scales. Recent work has demonstrated that Twitter data correlate with ILI rates at the municipal level )] TJ ET 0.267 0.267 0.267 rg BT 461.929 606.722 Td /F4 8.7 Tf [(23)] TJ ET 0.271 0.267 0.267 rg BT 471.566 609.103 Td /F1 8.7 Tf [(,)] TJ ET 0.267 0.267 0.267 rg BT 473.975 606.722 Td /F4 8.7 Tf [(24)] TJ ET 0.271 0.267 0.267 rg BT 483.613 605.214 Td /F1 9.8 Tf [( , suggesting that )] TJ ET BT 26.250 593.310 Td /F1 9.8 Tf [(Web data could improve forecasts for cities as well. More sophisticated models are typically used in practice \(18, 25\), and our )] TJ ET BT 26.250 581.405 Td /F1 9.8 Tf [(encouraging preliminary results motivate the need for experimenting with Twitter and GFT data in richer models, such as those )] TJ ET BT 26.250 569.500 Td /F1 9.8 Tf [(that take full advantage of variables unique to social media \(e.g., daily, rather than weekly, ILI estimates\).)] TJ ET BT 26.250 532.898 Td /F4 12.0 Tf [(Competing Interests)] TJ ET BT 26.250 512.943 Td /F1 9.8 Tf [(Dr. Dredze reports receipt of compensation for travel for talks at various academic, corporate, and governmental entities and )] TJ ET BT 26.250 501.039 Td /F1 9.8 Tf [(consulting for Directing Medicine, Progeny Systems, and Sickweather. Mr. Paul serves on the advisory board for Sickweather.)] TJ ET BT 26.250 464.436 Td /F4 12.0 Tf [(References)] TJ ET BT 26.250 436.982 Td /F1 9.8 Tf [(1.)] TJ ET BT 38.132 436.982 Td /F1 9.8 Tf [(Chretien JP, George D, Shaman J, Chitale RA, McKenzie FE. Influenza forecasting in human populations: a scoping review. )] TJ ET BT 26.250 425.077 Td /F1 9.8 Tf [(PLoS One. 2014;9\(4\):e94130. PubMed PMID:24714027.)] TJ ET BT 26.250 405.672 Td /F1 9.8 Tf [(2.)] TJ ET BT 38.132 405.672 Td /F1 9.8 Tf [(Nsoesie E, Mararthe M, Brownstein J. Forecasting peaks of seasonal influenza epidemics. PLoS Curr. 2013 Jun 21;5. )] TJ ET BT 26.250 393.768 Td /F1 9.8 Tf [(PubMed PMID:23873050.)] TJ ET BT 26.250 374.363 Td /F1 9.8 Tf [(3.)] TJ ET BT 38.132 374.363 Td /F1 9.8 Tf [(Shaman J, Karspeck A, Yang W, Tamerius J, Lipsitch M. Real-time influenza forecasts during the 2012-2013 season. Nat )] TJ ET BT 26.250 362.458 Td /F1 9.8 Tf [(Commun. 2013;4:2837. PubMed PMID:24302074.)] TJ ET BT 26.250 343.053 Td /F1 9.8 Tf [(4.)] TJ ET BT 38.132 343.053 Td /F1 9.8 Tf [(Soebiyanto RP, Adimi F, Kiang RK. Modeling and predicting seasonal influenza transmission in warm regions using )] TJ ET BT 26.250 331.149 Td /F1 9.8 Tf [(climatological parameters. PLoS One. 2010 Mar 1;5\(3\):e9450. PubMed PMID:20209164.)] TJ ET BT 26.250 311.744 Td /F1 9.8 Tf [(5.)] TJ ET BT 38.132 311.744 Td /F1 9.8 Tf [(Culotta, A. Towards detecting influenza epidemics by analyzing Twitter messages. In ACM Workshop on Social Media )] TJ ET BT 26.250 299.839 Td /F1 9.8 Tf [(Analytics. 2010.)] TJ ET BT 26.250 280.434 Td /F1 9.8 Tf [(6.)] TJ ET BT 38.132 280.434 Td /F1 9.8 Tf [(Paul, MJ, Dredze, M. You Are What You Tweet: Analyzing Twitter for Public Health. In International Conference on Weblogs )] TJ ET BT 26.250 268.530 Td /F1 9.8 Tf [(and Social Media \(ICWSM\). 2011.)] TJ ET BT 26.250 249.125 Td /F1 9.8 Tf [(7.)] TJ ET BT 38.132 249.125 Td /F1 9.8 Tf [(Lampos V, Cristianini N. Nowcasting Events from the Social Web with Statistical Learning. ACM Transactions on Intelligent )] TJ ET BT 26.250 237.220 Td /F1 9.8 Tf [(Systems and Technology; 2012 Sep 1;3\(4\):122. DOI: 10.1145/2337542.2337557)] TJ ET BT 26.250 217.815 Td /F1 9.8 Tf [(8.)] TJ ET BT 38.132 217.815 Td /F1 9.8 Tf [(Dredze, M. How Social Media Will Change Public Health. IEEE Intelligent Systems; vol. 27, no. 4, pp. 8184, Jul. 2012.)] TJ ET BT 26.250 198.411 Td /F1 9.8 Tf [(9.)] TJ ET BT 38.132 198.411 Td /F1 9.8 Tf [(Chew C, Eysenbach G. Pandemics in the age of Twitter: content analysis of Tweets during the 2009 H1N1 outbreak. PLoS )] TJ ET BT 26.250 186.506 Td /F1 9.8 Tf [(One. 2010 Nov 29;5\(11\):e14118. PubMed PMID:21124761.)] TJ ET BT 26.250 167.101 Td /F1 9.8 Tf [(10.)] TJ ET BT 43.553 167.101 Td /F1 9.8 Tf [(Salath M, Khandelwal S. Assessing vaccination sentiments with online social media: implications for infectious disease )] TJ ET BT 26.250 155.196 Td /F1 9.8 Tf [(dynamics and control. PLoS Comput Biol. 2011 Oct;7\(10\):e1002199. PubMed PMID:22022249.)] TJ ET BT 26.250 135.792 Td /F1 9.8 Tf [(11.)] TJ ET BT 43.553 135.792 Td /F1 9.8 Tf [(Lamb, A, Paul, MJ, Dredze, M. Separating Fact from Fear: Tracking Flu Infections on Twitter. In North American Chapter of )] TJ ET BT 26.250 123.887 Td /F1 9.8 Tf [(the Association for Computational Linguistics \(NAACL\). 2013.)] TJ ET BT 26.250 104.482 Td /F1 9.8 Tf [(12.)] TJ ET BT 43.553 104.482 Td /F1 9.8 Tf [(Gesualdo F, Stilo G, Agricola E, Gonfiantini MV, Pandolfi E, Velardi P, Tozzi AE. Influenza-like illness surveillance on )] TJ ET BT 26.250 92.577 Td /F1 9.8 Tf [(Twitter through automated learning of nave language. PLoS One. 2013;8\(12\):e82489. PubMed PMID:24324799.)] TJ ET BT 26.250 73.173 Td /F1 9.8 Tf [(13.)] TJ ET BT 43.553 73.173 Td /F1 9.8 Tf [(Ginsberg J, Mohebbi MH, Patel RS, Brammer L, Smolinski MS, Brilliant L. Detecting influenza epidemics using search )] TJ ET BT 26.250 61.268 Td /F1 9.8 Tf [(engine query data. Nature. 2009 Feb 19;457\(7232\):1012-4. PubMed PMID:19020500.)] TJ ET Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(7)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj 397 0 obj << /Type /Annot /Subtype /Link /A 398 0 R /Border [0 0 0] /H /I /Rect [ 446.7577 744.3722 456.3951 753.1905 ] >> endobj 398 0 obj << /Type /Action >> endobj 399 0 obj << /Type /Annot /Subtype /Link /A 400 0 R /Border [0 0 0] /H /I /Rect [ 458.8044 744.3722 468.4418 753.1905 ] >> endobj 400 0 obj << /Type /Action >> endobj 401 0 obj << /Type /Annot /Subtype /Link /A 402 0 R /Border [0 0 0] /H /I /Rect [ 363.3173 696.7532 372.9546 705.5715 ] >> endobj 402 0 obj << /Type /Action >> endobj 403 0 obj << /Type /Annot /Subtype /Link /A 404 0 R /Border [0 0 0] /H /I /Rect [ 296.6662 661.0389 306.3036 669.8573 ] >> endobj 404 0 obj << /Type /Action >> endobj 405 0 obj << /Type /Annot /Subtype /Link /A 406 0 R /Border [0 0 0] /H /I /Rect [ 235.9725 637.2294 245.6098 646.0477 ] >> endobj 406 0 obj << /Type /Action >> endobj 407 0 obj << /Type /Annot /Subtype /Link /A 408 0 R /Border [0 0 0] /H /I /Rect [ 461.9287 605.9199 471.5661 614.7382 ] >> endobj 408 0 obj << /Type /Action >> endobj 409 0 obj << /Type /Annot /Subtype /Link /A 410 0 R /Border [0 0 0] /H /I /Rect [ 473.9754 605.9199 483.6128 614.7382 ] >> endobj 410 0 obj << /Type /Action >> endobj 411 0 obj << /Type /Annot /Subtype /Link /A 412 0 R /Border [0 0 0] /H /I /Rect [ 446.7577 744.3722 456.3951 753.1905 ] >> endobj 412 0 obj << /Type /Action >> endobj 413 0 obj << /Type /Annot /Subtype /Link /A 414 0 R /Border [0 0 0] /H /I /Rect [ 458.8044 744.3722 468.4418 753.1905 ] >> endobj 414 0 obj << /Type /Action >> endobj 415 0 obj << /Type /Annot /Subtype /Link /A 416 0 R /Border [0 0 0] /H /I /Rect [ 363.3173 696.7532 372.9546 705.5715 ] >> endobj 416 0 obj << /Type /Action >> endobj 417 0 obj << /Type /Annot /Subtype /Link /A 418 0 R /Border [0 0 0] /H /I /Rect [ 296.6662 661.0389 306.3036 669.8573 ] >> endobj 418 0 obj << /Type /Action >> endobj 419 0 obj << /Type /Annot /Subtype /Link /A 420 0 R /Border [0 0 0] /H /I /Rect [ 235.9725 637.2294 245.6098 646.0477 ] >> endobj 420 0 obj << /Type /Action >> endobj 421 0 obj << /Type /Annot /Subtype /Link /A 422 0 R /Border [0 0 0] /H /I /Rect [ 461.9287 605.9199 471.5661 614.7382 ] >> endobj 422 0 obj << /Type /Action >> endobj 423 0 obj << /Type /Annot /Subtype /Link /A 424 0 R /Border [0 0 0] /H /I /Rect [ 473.9754 605.9199 483.6128 614.7382 ] >> endobj 424 0 obj << /Type /Action >> endobj 425 0 obj << /Type /Annot /Subtype /Link /A 426 0 R /Border [0 0 0] /H /I /Rect [ 446.7577 744.3722 456.3951 753.1905 ] >> endobj 426 0 obj << /Type /Action >> endobj 427 0 obj << /Type /Annot /Subtype /Link /A 428 0 R /Border [0 0 0] /H /I /Rect [ 458.8044 744.3722 468.4418 753.1905 ] >> endobj 428 0 obj << /Type /Action >> endobj 429 0 obj << /Type /Annot /Subtype /Link /A 430 0 R /Border [0 0 0] /H /I /Rect [ 363.3173 696.7532 372.9546 705.5715 ] >> endobj 430 0 obj << /Type /Action >> endobj 431 0 obj << /Type /Annot /Subtype /Link /A 432 0 R /Border [0 0 0] /H /I /Rect [ 296.6662 661.0389 306.3036 669.8573 ] >> endobj 432 0 obj << /Type /Action >> endobj 433 0 obj << /Type /Annot /Subtype /Link /A 434 0 R /Border [0 0 0] /H /I /Rect [ 235.9725 637.2294 245.6098 646.0477 ] >> endobj 434 0 obj << /Type /Action >> endobj 435 0 obj << /Type /Annot /Subtype /Link /A 436 0 R /Border [0 0 0] /H /I /Rect [ 461.9287 605.9199 471.5661 614.7382 ] >> endobj 436 0 obj << /Type /Action >> endobj 437 0 obj << /Type /Annot /Subtype /Link /A 438 0 R /Border [0 0 0] /H /I /Rect [ 473.9754 605.9199 483.6128 614.7382 ] >> endobj 438 0 obj << /Type /Action >> endobj 439 0 obj << /Type /Page /Parent 3 0 R /Contents 440 0 R >> endobj 440 0 obj << /Length 18571 >> stream 0.271 0.267 0.267 rg q 15.000 162.715 577.500 614.285 re W n 0.271 0.267 0.267 rg BT 26.250 759.976 Td /F1 9.8 Tf [(14.)] TJ ET BT 43.553 759.976 Td /F1 9.8 Tf [(Polgreen PM, Chen Y, Pennock DM, Nelson FD. Using internet searches for influenza surveillance. Clin Infect Dis. 2008 )] TJ ET BT 26.250 748.071 Td /F1 9.8 Tf [(Dec 1;47\(11\):1443-8. PubMed PMID:18954267.)] TJ ET BT 26.250 728.667 Td /F1 9.8 Tf [(15.)] TJ ET BT 43.553 728.667 Td /F1 9.8 Tf [(Yuan Q, Nsoesie EO, Lv B, Peng G, Chunara R, Brownstein JS. Monitoring influenza epidemics in china with search query )] TJ ET BT 26.250 716.762 Td /F1 9.8 Tf [(from baidu. PLoS One. 2013;8\(5\):e64323. PubMed PMID:23750192.)] TJ ET BT 26.250 697.357 Td /F1 9.8 Tf [(16.)] TJ ET BT 43.553 697.357 Td /F1 9.8 Tf [(Chunara, R. Aman, S. Smolinski, M. Brownstein, JS. Flu Near You: An Online Self-reported Influenza Surveillance System )] TJ ET BT 26.250 685.452 Td /F1 9.8 Tf [(in the USA. Online Journal of Public Health Informatics. 2013 Mar;5\(1\))] TJ ET BT 26.250 666.048 Td /F1 9.8 Tf [(17.)] TJ ET BT 43.553 666.048 Td /F1 9.8 Tf [(Lazer D, Kennedy R, King G, Vespignani A. Big data. The parable of Google Flu: traps in big data analysis. Science. 2014 )] TJ ET BT 26.250 654.143 Td /F1 9.8 Tf [(Mar 14;343\(6176\):1203-5. PubMed PMID:24626916.)] TJ ET BT 26.250 634.738 Td /F1 9.8 Tf [(18.)] TJ ET BT 43.553 634.738 Td /F1 9.8 Tf [(Shaman J, Karspeck A. Forecasting seasonal outbreaks of influenza. Proc Natl Acad Sci U S A. 2012 Dec )] TJ ET BT 26.250 622.833 Td /F1 9.8 Tf [(11;109\(50\):20425-30. PubMed PMID:23184969.)] TJ ET BT 26.250 603.429 Td /F1 9.8 Tf [(19.)] TJ ET BT 43.553 603.429 Td /F1 9.8 Tf [(Dugas AF, Jalalpour M, Gel Y, Levin S, Torcaso F, Igusa T, Rothman RE. Influenza forecasting with Google Flu Trends. )] TJ ET BT 26.250 591.524 Td /F1 9.8 Tf [(PLoS One. 2013;8\(2\):e56176. PubMed PMID:23457520.)] TJ ET BT 26.250 572.119 Td /F1 9.8 Tf [(20.)] TJ ET BT 43.553 572.119 Td /F1 9.8 Tf [(Santillana M, Zhang DW, Althouse BM, Ayers JW. What Can Digital Disease Detection Learn from \(an External Revision to\) )] TJ ET BT 26.250 560.214 Td /F1 9.8 Tf [(Google Flu Trends? Am J Prev Med. 2014 Sep;47\(3\):341-7. PubMed PMID:24997572.)] TJ ET BT 26.250 540.810 Td /F1 9.8 Tf [(21.)] TJ ET BT 43.553 540.810 Td /F1 9.8 Tf [(Copeland, P. Romano, R. Zhang, T. Hecht, G. Zigmond, D. Stefansen, C. Google Disease Trends: An Update. In )] TJ ET BT 26.250 528.905 Td /F1 9.8 Tf [(International Society of Neglected Tropical Diseases. 2013.)] TJ ET BT 26.250 509.500 Td /F1 9.8 Tf [(22.)] TJ ET BT 43.553 509.500 Td /F1 9.8 Tf [(Goel S, Hofman JM, Lahaie S, Pennock DM, Watts DJ. Predicting consumer behavior with Web search. Proc Natl Acad Sci )] TJ ET BT 26.250 497.595 Td /F1 9.8 Tf [(U S A. 2010 Oct 12;107\(41\):17486-90. PubMed PMID:20876140.)] TJ ET BT 26.250 478.191 Td /F1 9.8 Tf [(23.)] TJ ET BT 43.553 478.191 Td /F1 9.8 Tf [(Broniatowski DA, Paul MJ, Dredze M. National and local influenza surveillance through Twitter: an analysis of the 2012-)] TJ ET BT 26.250 466.286 Td /F1 9.8 Tf [(2013 influenza epidemic. PLoS One. 2013;8\(12\):e83672. PubMed PMID:24349542.)] TJ ET BT 26.250 446.881 Td /F1 9.8 Tf [(24.)] TJ ET BT 43.553 446.881 Td /F1 9.8 Tf [(Nagel AC, Tsou MH, Spitzberg BH, An L, Gawron JM, Gupta DK, Yang JA, Han S, Peddecord KM, Lindsay S, Sawyer MH. )] TJ ET BT 26.250 434.976 Td /F1 9.8 Tf [(The complex relationship of realspace events and messages in cyberspace: case study of influenza and pertussis using tweets. )] TJ ET BT 26.250 423.072 Td /F1 9.8 Tf [(J Med Internet Res. 2013 Oct 24;15\(10\):e237. PubMed PMID:24158773.)] TJ ET BT 26.250 403.667 Td /F1 9.8 Tf [(25.)] TJ ET BT 43.553 403.667 Td /F1 9.8 Tf [(Yang W, Karspeck A, Shaman J. Comparison of filtering methods for the modeling and retrospective forecasting of )] TJ ET BT 26.250 391.762 Td /F1 9.8 Tf [(influenza epidemics. PLoS Comput Biol. 2014 Apr;10\(4\):e1003583. PubMed PMID:24762780.)] TJ ET BT 26.250 372.357 Td /F1 9.8 Tf [(26.)] TJ ET BT 43.553 372.357 Td /F1 9.8 Tf [(Broniatowski DA, Paul MJ, Dredze M. Twitter: big data opportunities. Science. 2014 Jul 11;345\(6193\):148. PubMed )] TJ ET BT 26.250 360.453 Td /F1 9.8 Tf [(PMID:25013052.)] TJ ET BT 26.250 341.048 Td /F1 9.8 Tf [(27.)] TJ ET BT 43.553 341.048 Td /F1 9.8 Tf [(Dredze, M, Paul, M, Bergsma, S, Tran, H. Carmen: A Twitter Geolocation System with Applications to Public Health. In )] TJ ET BT 26.250 329.143 Td /F1 9.8 Tf [(AAAI Workshop on Expanding the Boundaries of Health Informatics Using Artificial Intelligence \(HIAI\). 2013.)] TJ ET BT 26.250 309.738 Td /F1 9.8 Tf [(28.)] TJ ET BT 43.553 309.738 Td /F1 9.8 Tf [(Chakraborty, P, Khadivi, P, Lewis, B, Mahendiran, A, Chen, J, Butler, P, Nsoesie, EO, Mekaru, SR, Brownstein, J, Marathe, )] TJ ET BT 26.250 297.834 Td /F1 9.8 Tf [(M, Ramakrishnan. Forecasting a Moving Target: Ensemble Models for ILI Case Count Predictions. In SIAM International )] TJ ET BT 26.250 285.929 Td /F1 9.8 Tf [(Conference on Data Mining. 2014)] TJ ET BT 26.250 266.524 Td /F1 9.8 Tf [(29.)] TJ ET BT 43.553 266.524 Td /F1 9.8 Tf [(Signorini, A, Polgreen, PM, Segre, AM. Using Twitter to estimate H1N1 influenza activity. In 9th Annual Conference of the )] TJ ET BT 26.250 254.619 Td /F1 9.8 Tf [(International Society for Disease Surveillance. 2010.)] TJ ET BT 26.250 235.215 Td /F1 9.8 Tf [(30.)] TJ ET BT 43.553 235.215 Td /F1 9.8 Tf [(Achrekar, H, Gandhe, A, Lazarus, R, Yu, S, Liu, B. Twitter Improves Seasonal Influenza Prediction. In International )] TJ ET BT 26.250 223.310 Td /F1 9.8 Tf [(Conference on Health Informatics, pp. 61-70. 2012.)] TJ ET BT 26.250 203.905 Td /F1 9.8 Tf [(31.)] TJ ET BT 43.553 203.905 Td /F1 9.8 Tf [(Nagar R, Yuan Q, Freifeld CC, Santillana M, Nojima A, Chunara R, Brownstein JS. A case study of the new york city 2012-)] TJ ET BT 26.250 192.000 Td /F1 9.8 Tf [(2013 influenza season with daily geocoded twitter data from temporal and spatiotemporal perspectives. J Med Internet Res. )] TJ ET BT 26.250 180.096 Td /F1 9.8 Tf [(2014 Oct 20;16\(10\):e236. PubMed PMID:25331122.)] TJ ET Q q 15.000 162.715 577.500 614.285 re W n 0.271 0.267 0.267 rg BT 26.250 759.976 Td /F1 9.8 Tf [(14.)] TJ ET BT 43.553 759.976 Td /F1 9.8 Tf [(Polgreen PM, Chen Y, Pennock DM, Nelson FD. Using internet searches for influenza surveillance. Clin Infect Dis. 2008 )] TJ ET BT 26.250 748.071 Td /F1 9.8 Tf [(Dec 1;47\(11\):1443-8. PubMed PMID:18954267.)] TJ ET BT 26.250 728.667 Td /F1 9.8 Tf [(15.)] TJ ET BT 43.553 728.667 Td /F1 9.8 Tf [(Yuan Q, Nsoesie EO, Lv B, Peng G, Chunara R, Brownstein JS. Monitoring influenza epidemics in china with search query )] TJ ET BT 26.250 716.762 Td /F1 9.8 Tf [(from baidu. PLoS One. 2013;8\(5\):e64323. PubMed PMID:23750192.)] TJ ET BT 26.250 697.357 Td /F1 9.8 Tf [(16.)] TJ ET BT 43.553 697.357 Td /F1 9.8 Tf [(Chunara, R. Aman, S. Smolinski, M. Brownstein, JS. Flu Near You: An Online Self-reported Influenza Surveillance System )] TJ ET BT 26.250 685.452 Td /F1 9.8 Tf [(in the USA. Online Journal of Public Health Informatics. 2013 Mar;5\(1\))] TJ ET BT 26.250 666.048 Td /F1 9.8 Tf [(17.)] TJ ET BT 43.553 666.048 Td /F1 9.8 Tf [(Lazer D, Kennedy R, King G, Vespignani A. Big data. The parable of Google Flu: traps in big data analysis. Science. 2014 )] TJ ET BT 26.250 654.143 Td /F1 9.8 Tf [(Mar 14;343\(6176\):1203-5. PubMed PMID:24626916.)] TJ ET BT 26.250 634.738 Td /F1 9.8 Tf [(18.)] TJ ET BT 43.553 634.738 Td /F1 9.8 Tf [(Shaman J, Karspeck A. Forecasting seasonal outbreaks of influenza. Proc Natl Acad Sci U S A. 2012 Dec )] TJ ET BT 26.250 622.833 Td /F1 9.8 Tf [(11;109\(50\):20425-30. PubMed PMID:23184969.)] TJ ET BT 26.250 603.429 Td /F1 9.8 Tf [(19.)] TJ ET BT 43.553 603.429 Td /F1 9.8 Tf [(Dugas AF, Jalalpour M, Gel Y, Levin S, Torcaso F, Igusa T, Rothman RE. Influenza forecasting with Google Flu Trends. )] TJ ET BT 26.250 591.524 Td /F1 9.8 Tf [(PLoS One. 2013;8\(2\):e56176. PubMed PMID:23457520.)] TJ ET BT 26.250 572.119 Td /F1 9.8 Tf [(20.)] TJ ET BT 43.553 572.119 Td /F1 9.8 Tf [(Santillana M, Zhang DW, Althouse BM, Ayers JW. What Can Digital Disease Detection Learn from \(an External Revision to\) )] TJ ET BT 26.250 560.214 Td /F1 9.8 Tf [(Google Flu Trends? Am J Prev Med. 2014 Sep;47\(3\):341-7. PubMed PMID:24997572.)] TJ ET BT 26.250 540.810 Td /F1 9.8 Tf [(21.)] TJ ET BT 43.553 540.810 Td /F1 9.8 Tf [(Copeland, P. Romano, R. Zhang, T. Hecht, G. Zigmond, D. Stefansen, C. Google Disease Trends: An Update. In )] TJ ET BT 26.250 528.905 Td /F1 9.8 Tf [(International Society of Neglected Tropical Diseases. 2013.)] TJ ET BT 26.250 509.500 Td /F1 9.8 Tf [(22.)] TJ ET BT 43.553 509.500 Td /F1 9.8 Tf [(Goel S, Hofman JM, Lahaie S, Pennock DM, Watts DJ. Predicting consumer behavior with Web search. Proc Natl Acad Sci )] TJ ET BT 26.250 497.595 Td /F1 9.8 Tf [(U S A. 2010 Oct 12;107\(41\):17486-90. PubMed PMID:20876140.)] TJ ET BT 26.250 478.191 Td /F1 9.8 Tf [(23.)] TJ ET BT 43.553 478.191 Td /F1 9.8 Tf [(Broniatowski DA, Paul MJ, Dredze M. National and local influenza surveillance through Twitter: an analysis of the 2012-)] TJ ET BT 26.250 466.286 Td /F1 9.8 Tf [(2013 influenza epidemic. PLoS One. 2013;8\(12\):e83672. PubMed PMID:24349542.)] TJ ET BT 26.250 446.881 Td /F1 9.8 Tf [(24.)] TJ ET BT 43.553 446.881 Td /F1 9.8 Tf [(Nagel AC, Tsou MH, Spitzberg BH, An L, Gawron JM, Gupta DK, Yang JA, Han S, Peddecord KM, Lindsay S, Sawyer MH. )] TJ ET BT 26.250 434.976 Td /F1 9.8 Tf [(The complex relationship of realspace events and messages in cyberspace: case study of influenza and pertussis using tweets. )] TJ ET BT 26.250 423.072 Td /F1 9.8 Tf [(J Med Internet Res. 2013 Oct 24;15\(10\):e237. PubMed PMID:24158773.)] TJ ET BT 26.250 403.667 Td /F1 9.8 Tf [(25.)] TJ ET BT 43.553 403.667 Td /F1 9.8 Tf [(Yang W, Karspeck A, Shaman J. Comparison of filtering methods for the modeling and retrospective forecasting of )] TJ ET BT 26.250 391.762 Td /F1 9.8 Tf [(influenza epidemics. PLoS Comput Biol. 2014 Apr;10\(4\):e1003583. PubMed PMID:24762780.)] TJ ET BT 26.250 372.357 Td /F1 9.8 Tf [(26.)] TJ ET BT 43.553 372.357 Td /F1 9.8 Tf [(Broniatowski DA, Paul MJ, Dredze M. Twitter: big data opportunities. Science. 2014 Jul 11;345\(6193\):148. PubMed )] TJ ET BT 26.250 360.453 Td /F1 9.8 Tf [(PMID:25013052.)] TJ ET BT 26.250 341.048 Td /F1 9.8 Tf [(27.)] TJ ET BT 43.553 341.048 Td /F1 9.8 Tf [(Dredze, M, Paul, M, Bergsma, S, Tran, H. Carmen: A Twitter Geolocation System with Applications to Public Health. In )] TJ ET BT 26.250 329.143 Td /F1 9.8 Tf [(AAAI Workshop on Expanding the Boundaries of Health Informatics Using Artificial Intelligence \(HIAI\). 2013.)] TJ ET BT 26.250 309.738 Td /F1 9.8 Tf [(28.)] TJ ET BT 43.553 309.738 Td /F1 9.8 Tf [(Chakraborty, P, Khadivi, P, Lewis, B, Mahendiran, A, Chen, J, Butler, P, Nsoesie, EO, Mekaru, SR, Brownstein, J, Marathe, )] TJ ET BT 26.250 297.834 Td /F1 9.8 Tf [(M, Ramakrishnan. Forecasting a Moving Target: Ensemble Models for ILI Case Count Predictions. In SIAM International )] TJ ET BT 26.250 285.929 Td /F1 9.8 Tf [(Conference on Data Mining. 2014)] TJ ET BT 26.250 266.524 Td /F1 9.8 Tf [(29.)] TJ ET BT 43.553 266.524 Td /F1 9.8 Tf [(Signorini, A, Polgreen, PM, Segre, AM. Using Twitter to estimate H1N1 influenza activity. In 9th Annual Conference of the )] TJ ET BT 26.250 254.619 Td /F1 9.8 Tf [(International Society for Disease Surveillance. 2010.)] TJ ET BT 26.250 235.215 Td /F1 9.8 Tf [(30.)] TJ ET BT 43.553 235.215 Td /F1 9.8 Tf [(Achrekar, H, Gandhe, A, Lazarus, R, Yu, S, Liu, B. Twitter Improves Seasonal Influenza Prediction. In International )] TJ ET BT 26.250 223.310 Td /F1 9.8 Tf [(Conference on Health Informatics, pp. 61-70. 2012.)] TJ ET BT 26.250 203.905 Td /F1 9.8 Tf [(31.)] TJ ET BT 43.553 203.905 Td /F1 9.8 Tf [(Nagar R, Yuan Q, Freifeld CC, Santillana M, Nojima A, Chunara R, Brownstein JS. A case study of the new york city 2012-)] TJ ET BT 26.250 192.000 Td /F1 9.8 Tf [(2013 influenza season with daily geocoded twitter data from temporal and spatiotemporal perspectives. J Med Internet Res. )] TJ ET BT 26.250 180.096 Td /F1 9.8 Tf [(2014 Oct 20;16\(10\):e236. PubMed PMID:25331122.)] TJ ET Q q 15.000 162.715 577.500 614.285 re W n 0.271 0.267 0.267 rg BT 26.250 759.976 Td /F1 9.8 Tf [(14.)] TJ ET BT 43.553 759.976 Td /F1 9.8 Tf [(Polgreen PM, Chen Y, Pennock DM, Nelson FD. Using internet searches for influenza surveillance. Clin Infect Dis. 2008 )] TJ ET BT 26.250 748.071 Td /F1 9.8 Tf [(Dec 1;47\(11\):1443-8. PubMed PMID:18954267.)] TJ ET BT 26.250 728.667 Td /F1 9.8 Tf [(15.)] TJ ET BT 43.553 728.667 Td /F1 9.8 Tf [(Yuan Q, Nsoesie EO, Lv B, Peng G, Chunara R, Brownstein JS. Monitoring influenza epidemics in china with search query )] TJ ET BT 26.250 716.762 Td /F1 9.8 Tf [(from baidu. PLoS One. 2013;8\(5\):e64323. PubMed PMID:23750192.)] TJ ET BT 26.250 697.357 Td /F1 9.8 Tf [(16.)] TJ ET BT 43.553 697.357 Td /F1 9.8 Tf [(Chunara, R. Aman, S. Smolinski, M. Brownstein, JS. Flu Near You: An Online Self-reported Influenza Surveillance System )] TJ ET BT 26.250 685.452 Td /F1 9.8 Tf [(in the USA. Online Journal of Public Health Informatics. 2013 Mar;5\(1\))] TJ ET BT 26.250 666.048 Td /F1 9.8 Tf [(17.)] TJ ET BT 43.553 666.048 Td /F1 9.8 Tf [(Lazer D, Kennedy R, King G, Vespignani A. Big data. The parable of Google Flu: traps in big data analysis. Science. 2014 )] TJ ET BT 26.250 654.143 Td /F1 9.8 Tf [(Mar 14;343\(6176\):1203-5. PubMed PMID:24626916.)] TJ ET BT 26.250 634.738 Td /F1 9.8 Tf [(18.)] TJ ET BT 43.553 634.738 Td /F1 9.8 Tf [(Shaman J, Karspeck A. Forecasting seasonal outbreaks of influenza. Proc Natl Acad Sci U S A. 2012 Dec )] TJ ET BT 26.250 622.833 Td /F1 9.8 Tf [(11;109\(50\):20425-30. PubMed PMID:23184969.)] TJ ET BT 26.250 603.429 Td /F1 9.8 Tf [(19.)] TJ ET BT 43.553 603.429 Td /F1 9.8 Tf [(Dugas AF, Jalalpour M, Gel Y, Levin S, Torcaso F, Igusa T, Rothman RE. Influenza forecasting with Google Flu Trends. )] TJ ET BT 26.250 591.524 Td /F1 9.8 Tf [(PLoS One. 2013;8\(2\):e56176. PubMed PMID:23457520.)] TJ ET BT 26.250 572.119 Td /F1 9.8 Tf [(20.)] TJ ET BT 43.553 572.119 Td /F1 9.8 Tf [(Santillana M, Zhang DW, Althouse BM, Ayers JW. What Can Digital Disease Detection Learn from \(an External Revision to\) )] TJ ET BT 26.250 560.214 Td /F1 9.8 Tf [(Google Flu Trends? Am J Prev Med. 2014 Sep;47\(3\):341-7. PubMed PMID:24997572.)] TJ ET BT 26.250 540.810 Td /F1 9.8 Tf [(21.)] TJ ET BT 43.553 540.810 Td /F1 9.8 Tf [(Copeland, P. Romano, R. Zhang, T. Hecht, G. Zigmond, D. Stefansen, C. Google Disease Trends: An Update. In )] TJ ET BT 26.250 528.905 Td /F1 9.8 Tf [(International Society of Neglected Tropical Diseases. 2013.)] TJ ET BT 26.250 509.500 Td /F1 9.8 Tf [(22.)] TJ ET BT 43.553 509.500 Td /F1 9.8 Tf [(Goel S, Hofman JM, Lahaie S, Pennock DM, Watts DJ. Predicting consumer behavior with Web search. Proc Natl Acad Sci )] TJ ET BT 26.250 497.595 Td /F1 9.8 Tf [(U S A. 2010 Oct 12;107\(41\):17486-90. PubMed PMID:20876140.)] TJ ET BT 26.250 478.191 Td /F1 9.8 Tf [(23.)] TJ ET BT 43.553 478.191 Td /F1 9.8 Tf [(Broniatowski DA, Paul MJ, Dredze M. National and local influenza surveillance through Twitter: an analysis of the 2012-)] TJ ET BT 26.250 466.286 Td /F1 9.8 Tf [(2013 influenza epidemic. PLoS One. 2013;8\(12\):e83672. PubMed PMID:24349542.)] TJ ET BT 26.250 446.881 Td /F1 9.8 Tf [(24.)] TJ ET BT 43.553 446.881 Td /F1 9.8 Tf [(Nagel AC, Tsou MH, Spitzberg BH, An L, Gawron JM, Gupta DK, Yang JA, Han S, Peddecord KM, Lindsay S, Sawyer MH. )] TJ ET BT 26.250 434.976 Td /F1 9.8 Tf [(The complex relationship of realspace events and messages in cyberspace: case study of influenza and pertussis using tweets. )] TJ ET BT 26.250 423.072 Td /F1 9.8 Tf [(J Med Internet Res. 2013 Oct 24;15\(10\):e237. PubMed PMID:24158773.)] TJ ET BT 26.250 403.667 Td /F1 9.8 Tf [(25.)] TJ ET BT 43.553 403.667 Td /F1 9.8 Tf [(Yang W, Karspeck A, Shaman J. Comparison of filtering methods for the modeling and retrospective forecasting of )] TJ ET BT 26.250 391.762 Td /F1 9.8 Tf [(influenza epidemics. PLoS Comput Biol. 2014 Apr;10\(4\):e1003583. PubMed PMID:24762780.)] TJ ET BT 26.250 372.357 Td /F1 9.8 Tf [(26.)] TJ ET BT 43.553 372.357 Td /F1 9.8 Tf [(Broniatowski DA, Paul MJ, Dredze M. Twitter: big data opportunities. Science. 2014 Jul 11;345\(6193\):148. PubMed )] TJ ET BT 26.250 360.453 Td /F1 9.8 Tf [(PMID:25013052.)] TJ ET BT 26.250 341.048 Td /F1 9.8 Tf [(27.)] TJ ET BT 43.553 341.048 Td /F1 9.8 Tf [(Dredze, M, Paul, M, Bergsma, S, Tran, H. Carmen: A Twitter Geolocation System with Applications to Public Health. In )] TJ ET BT 26.250 329.143 Td /F1 9.8 Tf [(AAAI Workshop on Expanding the Boundaries of Health Informatics Using Artificial Intelligence \(HIAI\). 2013.)] TJ ET BT 26.250 309.738 Td /F1 9.8 Tf [(28.)] TJ ET BT 43.553 309.738 Td /F1 9.8 Tf [(Chakraborty, P, Khadivi, P, Lewis, B, Mahendiran, A, Chen, J, Butler, P, Nsoesie, EO, Mekaru, SR, Brownstein, J, Marathe, )] TJ ET BT 26.250 297.834 Td /F1 9.8 Tf [(M, Ramakrishnan. Forecasting a Moving Target: Ensemble Models for ILI Case Count Predictions. In SIAM International )] TJ ET BT 26.250 285.929 Td /F1 9.8 Tf [(Conference on Data Mining. 2014)] TJ ET BT 26.250 266.524 Td /F1 9.8 Tf [(29.)] TJ ET BT 43.553 266.524 Td /F1 9.8 Tf [(Signorini, A, Polgreen, PM, Segre, AM. Using Twitter to estimate H1N1 influenza activity. In 9th Annual Conference of the )] TJ ET BT 26.250 254.619 Td /F1 9.8 Tf [(International Society for Disease Surveillance. 2010.)] TJ ET BT 26.250 235.215 Td /F1 9.8 Tf [(30.)] TJ ET BT 43.553 235.215 Td /F1 9.8 Tf [(Achrekar, H, Gandhe, A, Lazarus, R, Yu, S, Liu, B. Twitter Improves Seasonal Influenza Prediction. In International )] TJ ET BT 26.250 223.310 Td /F1 9.8 Tf [(Conference on Health Informatics, pp. 61-70. 2012.)] TJ ET BT 26.250 203.905 Td /F1 9.8 Tf [(31.)] TJ ET BT 43.553 203.905 Td /F1 9.8 Tf [(Nagar R, Yuan Q, Freifeld CC, Santillana M, Nojima A, Chunara R, Brownstein JS. A case study of the new york city 2012-)] TJ ET BT 26.250 192.000 Td /F1 9.8 Tf [(2013 influenza season with daily geocoded twitter data from temporal and spatiotemporal perspectives. J Med Internet Res. )] TJ ET BT 26.250 180.096 Td /F1 9.8 Tf [(2014 Oct 20;16\(10\):e236. PubMed PMID:25331122.)] TJ ET Q q 0.000 0.000 0.000 rg BT 291.710 19.825 Td /F1 11.0 Tf [(8)] TJ ET BT 25.000 19.825 Td /F1 11.0 Tf [(PLOS Currents Outbreaks)] TJ ET Q endstream endobj xref 0 441 0000000000 65535 f 0000000008 00000 n 0000000073 00000 n 0000000119 00000 n 0000001158 00000 n 0000001195 00000 n 0000001383 00000 n 0000002303 00000 n 0000037194 00000 n 0000037301 00000 n 0000037409 00000 n 0000037520 00000 n 0000037633 00000 n 0000038021 00000 n 0000047689 00000 n 0000047816 00000 n 0000047950 00000 n 0000048077 00000 n 0000048198 00000 n 0000048324 00000 n 0000048434 00000 n 0000048561 00000 n 0000048663 00000 n 0000048791 00000 n 0000048913 00000 n 0000049041 00000 n 0000049077 00000 n 0000049205 00000 n 0000049241 00000 n 0000049369 00000 n 0000049405 00000 n 0000049533 00000 n 0000049569 00000 n 0000049697 00000 n 0000049733 00000 n 0000049861 00000 n 0000049897 00000 n 0000050025 00000 n 0000050061 00000 n 0000050189 00000 n 0000050225 00000 n 0000050353 00000 n 0000050389 00000 n 0000050517 00000 n 0000050553 00000 n 0000050681 00000 n 0000050717 00000 n 0000050845 00000 n 0000050881 00000 n 0000051009 00000 n 0000051045 00000 n 0000051173 00000 n 0000051209 00000 n 0000051337 00000 n 0000051373 00000 n 0000051501 00000 n 0000051537 00000 n 0000051665 00000 n 0000051701 00000 n 0000051829 00000 n 0000051865 00000 n 0000051993 00000 n 0000052029 00000 n 0000052157 00000 n 0000052193 00000 n 0000052321 00000 n 0000052357 00000 n 0000052485 00000 n 0000052521 00000 n 0000052649 00000 n 0000052685 00000 n 0000052813 00000 n 0000052849 00000 n 0000052977 00000 n 0000053013 00000 n 0000053141 00000 n 0000053177 00000 n 0000053305 00000 n 0000053341 00000 n 0000053469 00000 n 0000053505 00000 n 0000053633 00000 n 0000053669 00000 n 0000053797 00000 n 0000053833 00000 n 0000053961 00000 n 0000053997 00000 n 0000054123 00000 n 0000054159 00000 n 0000054286 00000 n 0000054420 00000 n 0000054547 00000 n 0000054668 00000 n 0000054794 00000 n 0000054904 00000 n 0000055031 00000 n 0000055133 00000 n 0000055261 00000 n 0000055383 00000 n 0000055511 00000 n 0000055547 00000 n 0000055677 00000 n 0000055714 00000 n 0000055844 00000 n 0000055881 00000 n 0000056011 00000 n 0000056048 00000 n 0000056178 00000 n 0000056215 00000 n 0000056345 00000 n 0000056382 00000 n 0000056512 00000 n 0000056549 00000 n 0000056679 00000 n 0000056716 00000 n 0000056846 00000 n 0000056883 00000 n 0000057013 00000 n 0000057050 00000 n 0000057180 00000 n 0000057217 00000 n 0000057347 00000 n 0000057384 00000 n 0000057514 00000 n 0000057551 00000 n 0000057681 00000 n 0000057718 00000 n 0000057848 00000 n 0000057885 00000 n 0000058015 00000 n 0000058052 00000 n 0000058182 00000 n 0000058219 00000 n 0000058349 00000 n 0000058386 00000 n 0000058516 00000 n 0000058553 00000 n 0000058683 00000 n 0000058720 00000 n 0000058850 00000 n 0000058887 00000 n 0000059017 00000 n 0000059054 00000 n 0000059184 00000 n 0000059221 00000 n 0000059351 00000 n 0000059388 00000 n 0000059518 00000 n 0000059555 00000 n 0000059685 00000 n 0000059722 00000 n 0000059852 00000 n 0000059889 00000 n 0000060019 00000 n 0000060056 00000 n 0000060186 00000 n 0000060223 00000 n 0000060353 00000 n 0000060390 00000 n 0000060520 00000 n 0000060557 00000 n 0000060685 00000 n 0000060722 00000 n 0000060851 00000 n 0000060986 00000 n 0000061115 00000 n 0000061237 00000 n 0000061365 00000 n 0000061476 00000 n 0000061605 00000 n 0000061708 00000 n 0000061838 00000 n 0000061961 00000 n 0000062091 00000 n 0000062128 00000 n 0000062258 00000 n 0000062295 00000 n 0000062425 00000 n 0000062462 00000 n 0000062592 00000 n 0000062629 00000 n 0000062759 00000 n 0000062796 00000 n 0000062926 00000 n 0000062963 00000 n 0000063093 00000 n 0000063130 00000 n 0000063260 00000 n 0000063297 00000 n 0000063427 00000 n 0000063464 00000 n 0000063594 00000 n 0000063631 00000 n 0000063761 00000 n 0000063798 00000 n 0000063928 00000 n 0000063965 00000 n 0000064095 00000 n 0000064132 00000 n 0000064262 00000 n 0000064299 00000 n 0000064429 00000 n 0000064466 00000 n 0000064596 00000 n 0000064633 00000 n 0000064763 00000 n 0000064800 00000 n 0000064930 00000 n 0000064967 00000 n 0000065097 00000 n 0000065134 00000 n 0000065264 00000 n 0000065301 00000 n 0000065431 00000 n 0000065468 00000 n 0000065598 00000 n 0000065635 00000 n 0000065765 00000 n 0000065802 00000 n 0000065932 00000 n 0000065969 00000 n 0000066099 00000 n 0000066136 00000 n 0000066266 00000 n 0000066303 00000 n 0000066433 00000 n 0000066470 00000 n 0000066600 00000 n 0000066637 00000 n 0000066767 00000 n 0000066804 00000 n 0000066934 00000 n 0000066971 00000 n 0000067101 00000 n 0000067138 00000 n 0000067266 00000 n 0000067303 00000 n 0000067622 00000 n 0000094345 00000 n 0000094462 00000 n 0000094591 00000 n 0000094700 00000 n 0000094830 00000 n 0000094867 00000 n 0000094997 00000 n 0000095034 00000 n 0000095164 00000 n 0000095201 00000 n 0000095331 00000 n 0000095368 00000 n 0000095497 00000 n 0000095593 00000 n 0000095723 00000 n 0000095760 00000 n 0000095890 00000 n 0000095927 00000 n 0000096057 00000 n 0000096094 00000 n 0000097621 00000 n 0000097917 00000 n 0000098047 00000 n 0000098084 00000 n 0000098433 00000 n 0000098702 00000 n 0000099093 00000 n 0000099364 00000 n 0000099755 00000 n 0000100026 00000 n 0000102184 00000 n 0000102484 00000 n 0000103170 00000 n 0000103452 00000 n 0000103581 00000 n 0000103690 00000 n 0000103820 00000 n 0000103857 00000 n 0000103987 00000 n 0000104024 00000 n 0000104154 00000 n 0000104191 00000 n 0000104321 00000 n 0000104358 00000 n 0000104487 00000 n 0000104583 00000 n 0000104713 00000 n 0000104750 00000 n 0000104880 00000 n 0000104917 00000 n 0000105047 00000 n 0000105084 00000 n 0000106611 00000 n 0000106907 00000 n 0000107037 00000 n 0000107074 00000 n 0000107423 00000 n 0000107692 00000 n 0000108083 00000 n 0000108354 00000 n 0000108745 00000 n 0000109016 00000 n 0000111174 00000 n 0000111474 00000 n 0000112160 00000 n 0000112442 00000 n 0000112571 00000 n 0000112680 00000 n 0000112810 00000 n 0000112847 00000 n 0000112977 00000 n 0000113014 00000 n 0000113144 00000 n 0000113181 00000 n 0000113311 00000 n 0000113348 00000 n 0000113477 00000 n 0000113573 00000 n 0000113703 00000 n 0000113740 00000 n 0000113870 00000 n 0000113907 00000 n 0000114037 00000 n 0000114074 00000 n 0000115601 00000 n 0000115897 00000 n 0000116027 00000 n 0000116064 00000 n 0000116413 00000 n 0000116682 00000 n 0000117073 00000 n 0000117344 00000 n 0000117735 00000 n 0000118006 00000 n 0000120164 00000 n 0000120464 00000 n 0000121150 00000 n 0000121432 00000 n 0000121559 00000 n 0000159058 00000 n 0000161039 00000 n 0000161349 00000 n 0000161479 00000 n 0000161516 00000 n 0000161646 00000 n 0000161683 00000 n 0000163664 00000 n 0000163974 00000 n 0000164104 00000 n 0000164141 00000 n 0000164271 00000 n 0000164308 00000 n 0000166289 00000 n 0000166599 00000 n 0000166729 00000 n 0000166766 00000 n 0000166896 00000 n 0000166933 00000 n 0000167000 00000 n 0000198601 00000 n 0000198704 00000 n 0000224991 00000 n 0000225120 00000 n 0000225240 00000 n 0000226157 00000 n 0000283789 00000 n 0000283918 00000 n 0000284038 00000 n 0000284955 00000 n 0000342587 00000 n 0000342716 00000 n 0000342836 00000 n 0000343753 00000 n 0000401385 00000 n 0000402302 00000 n 0000459934 00000 n 0000460061 00000 n 0000471276 00000 n 0000471405 00000 n 0000471518 00000 n 0000472435 00000 n 0000515520 00000 n 0000515650 00000 n 0000515687 00000 n 0000515816 00000 n 0000515929 00000 n 0000516846 00000 n 0000559931 00000 n 0000560061 00000 n 0000560098 00000 n 0000560227 00000 n 0000560340 00000 n 0000561257 00000 n 0000604342 00000 n 0000604472 00000 n 0000604509 00000 n 0000605426 00000 n 0000648511 00000 n 0000648758 00000 n 0000673504 00000 n 0000673634 00000 n 0000673671 00000 n 0000673801 00000 n 0000673838 00000 n 0000673968 00000 n 0000674005 00000 n 0000674135 00000 n 0000674172 00000 n 0000674302 00000 n 0000674339 00000 n 0000674469 00000 n 0000674506 00000 n 0000674636 00000 n 0000674673 00000 n 0000674803 00000 n 0000674840 00000 n 0000674970 00000 n 0000675007 00000 n 0000675137 00000 n 0000675174 00000 n 0000675304 00000 n 0000675341 00000 n 0000675471 00000 n 0000675508 00000 n 0000675638 00000 n 0000675675 00000 n 0000675805 00000 n 0000675842 00000 n 0000675972 00000 n 0000676009 00000 n 0000676139 00000 n 0000676176 00000 n 0000676306 00000 n 0000676343 00000 n 0000676473 00000 n 0000676510 00000 n 0000676640 00000 n 0000676677 00000 n 0000676807 00000 n 0000676844 00000 n 0000676974 00000 n 0000677011 00000 n 0000677078 00000 n trailer << /Size 441 /Root 1 0 R /Info 5 0 R >> startxref 695704 %%EOF