Skip to main content
. 2019 Aug 2;15(8):e1007258. doi: 10.1371/journal.pcbi.1007258

Table 3. Mean squared error of near-term forecasts.

ILIp was generated with ILIp alone and ILIp+GFT by appending corrected GFT to ILIp. The lower error in each row is underscored. P-values from a paired Wilcoxon signed rank test that the median of error in ILIp forecasts are greater than the median of errors in ILIp+GFT forecasts are also shown; cases where p > .05 are denoted by an asterisk (*).

ILIp ILIp + GFT p
Overall 0.761 0.605 < .001
1 week ahead 0.327 0.294 .01
2 week ahead 0.611 0.459 < .001
3 week ahead 0.907 0.700 < .001
4 week ahead 1.199 0.968 < .001
National 0.452 0.367 < .001
Region 1 0.248 0.221 < .001
Region 2 0.605 0.503 .01
Region 3* 0.857 0.764 .25
Region 4 0.969 0.685 .03
Region 5* 0.526 0.582 .67
Region 6 1.554 1.233 < .001
Region 7 1.013 0.810 < .001
Region 8 0.445 0.316 < .001
Region 9 1.149 0.608 < .001
Region 10* 0.553 0.568 .76
2010/11 0.747 0.549 < .001
2011/12 0.268 0.229 < .001
2012/13* 1.035 0.866 .24
2013/14 0.643 0.549 < .001
2014/15 1.091 0.819 < .001