Skip to main content
. 2021 Aug 27;12:5173. doi: 10.1038/s41467-021-25207-0

Table 2.

Detailed summary of forecast evaluation for Poland (based on ECDC data). C0.5 and C0.95 denote coverage rates of the 50% and 95% prediction intervals; AE and WIS stand for the mean absolute error and mean-weighted interval score.

Poland, cases
1-week ahead incident 2-week ahead incident 1-week ahead cumulative 2-week ahead cumulative
Model AE WIS C0.5 C0.95 AE WIS C0.5 C0.95 AE WIS C0.5 C0.95 AE WIS C0.5 C0.95
epiforecasts-EpiExpert 13,643 9,574 7/10 8/10 37,395 24,980 2/9 5/9 13,620 9,596 7/10 8/10 52,523 33,602 2/9 6/9
epiforecasts-EpiNow2 11,006 7,041 5/10 7/10 38,906 25,308 2/9 6/9 11,028 7,048 5/10 7/10 47,373 30,303 2/9 7/9
ICM-agentModel 2/4 4/4 0/3 1/3 1/4 3/4 0/3 2/3
ITWW-county_repro 18,149 14,687 2/10 4/10 33,298 27,208 2/9 4/9 17,227 13,786 3/10 5/10 50,638 41,115 1/9 4/9
LANL-GrowthRate 15,956a 9,490a 3/7 7/7 49,295a 27,220a 1/6 6/6 15,269 9,311 3/10 10/10 62,801 36,562 2/9 8/9
MIMUW-StochSEIR 3/5 5/5 2/4 2/4 2/5 5/5 1/4 2/4
MIT_CovidAnalytics-DELPHI 32,620a 23,266a 2/9 5/9 60,490a 45,815a 1/8 4/8
MOCOS-agent1 13,273 9,124 2/10 5/10 31,610 24,976 0/9 3/9 13,273 9,124 2/10 5/10 43,215 32,106 1/9 3/9
SDSC_ISG-TrendModel 7,633 7,656
USC-SIkJalpha 10,292 0/1 1/1 24,138 13,560 0/1 1/1 35,390
KIT-baseline 28,164 18,119 5/10 9/10 52,890 35,107 2/9 6/9 28,235 18,154 5/10 9/10 80,765 53,656 2/9 6/9
KIT-extrapolation_baseline 18,311 11,917 6/10 10/10 55,060 34,212 3/9 7/9 18,289 11,912 6/10 10/10 76,607 45,935 3/9 8/9
KIT-time_series_baseline 22,497 14,100 5/10 10/10 60,079 37,980 5/9 9/9 22,475 14,098 5/10 10/10 84,530 52,303 4/9 9/9
KITCOVIDhub-inverse_wis_ensemble 12,768 8,456 4/10 9/10 36,229 24,628 3/9 6/9 11,865 7,733 4/10 9/10 44,477 30,643 2/9 6/9
KITCOVIDhub-mean_ensemble 12,982 8,320 3/10 9/10 36,338 23,598 3/9 6/9 12,051 7,582 5/10 10/10 44,254 29,461 2/9 7/9
KITCOVIDhub-median_ensemble 14,196 8,862 5/10 9/10 39,829 24,620 2/9 6/9 14,033 8,698 5/10 9/10 50,935 30,699 2/9 5/9
Poland, deaths
1-week ahead incident 2-week ahead incident 1-week ahead cumulative 2-week ahead cumulative
Model AE WIS C0.5 C0.95 AE WIS C0.5 C0.95 AE WIS C0.5 C0.95 AE WIS C0.5 C0.95
epiforecasts-EpiExpert 285 176 4/10 10/10 605 374 1/9 8/9 285 176 4/10 10/10 874 530 3/9 8/9
epiforecasts-EpiNow2 386 261 3/10 7/10 1,110 781 2/9 4/9 386 261 3/10 7/10 1,528 1,049 2/9 5/9
ICM-agentModel 752a 672a 2/8 3/8 1,881a 1,237a 0/7 2/7 1,178a 727a 1/8 4/8 2,955a 1,865a 0/7 3/7
Imperial-ensemble2 397 238 3/10 7/10 369 211 3/10 8/10
ITWW-county_repro 525 484 0/10 1/10 701 613 0/9 2/9 524 483 0/10 1/10 1,219 1,085 0/9 2/9
LANL-GrowthRate 239a 175a 4/7 7/7 404a 251a 3/6 6/6 216 152 5/10 10/10 637 404 3/9 8/9
MIMUW-StochSEIR 1/5 4/5 1/4 2/4 1/5 4/5 0/4 4/4
MIT_CovidAnalytics-DELPHI 512a 329a 2/9 5/9 663a 434a 1/8 6/8 597a 417a 2/9 6/9 1,075a 782a 1/8 3/8
MOCOS-agent1 194 147 9/10 10/10 420 272 7/9 8/9 194 147 9/10 10/10 556 382 7/9 9/9
SDSC_ISG-TrendModel 154 154
USC-SIkJalpha 206 0/1 1/1 240 256 0/1 0/1 242
KIT-baseline 437 275 5/10 10/10 834 529 2/9 7/9 437 274 5/10 10/10 1,245 793 2/9 6/9
KIT-extrapolation_baseline 408 286 6/10 8/10 996 702 5/9 7/9 408 286 6/10 8/10 1,423 989 4/9 7/9
KIT-time_series_baseline 546 339 6/10 10/10 1,371 856 5/9 8/9 546 339 6/10 10/10 1,921 1,212 4/9 8/9
KITCOVIDhub-inverse_wis_ensemble 220 153 6/10 10/10 488 313 4/9 8/9 242 162 7/10 10/10 702 450 4/9 9/9
KITCOVIDhub-mean_ensemble 252 163 7/10 9/10 585 362 4/9 8/9 265 171 7/10 9/10 815 522 4/9 9/9
KITCOVIDhub-median_ensemble 215 148 6/10 10/10 471 289 2/9 9/9 231 160 6/10 10/10 707 458 4/9 9/9

aEntries where scores were imputed for at least 1 week. Weighted interval scores and absolute errors were imputed with the worst (largest) score achieved by any other forecast for the respective target and week. Models marked thus received a pessimistic assessment of their performance. If a model covered less than two-thirds of the evaluation period, results are omitted.