Extended Data Table 3.
Results of the external predictive validity (out-of-sample validation) of the estimated no-pandemic counterfactual weekly deaths from the ensemble of Bayesian models. Each number represents the total error over the validation period, averaged across countries
| Validation year | Age group (years) | Projection error (relative projection error) | Absolute projection error (relative absolute projection error) | Percent covered by 95% credible interval | |||
|---|---|---|---|---|---|---|---|
| Men | Women | Men | Women | Men | Women | ||
| 2017 | 0-64 | 134 (2%) | 28 (1%) | 265 (7%) | 177(9%) | 96% | 98% |
| 65+ | 878 (5%) | 1201 (6%) | 1036 (7%) | 1343 (7%) | 93% | 91% | |
| 2018 | 0-64 | -92 (-3%) | -49 (-3%) | 278 (8%) | 186(9%) | 95% | 97% |
| 65+ | -223 (-2%) | -370 (-2%) | 1058(7%) | 1420 (8%) | 88% | 87% | |
| 2019 | 0-64 | 112(2%) | 50 (2%) | 261 (8%) | 164(9%) | 96% | 96% |
| 65+ | 430 (2%) | 558 (2%) | 696 (4%) | 898 (5%) | 98% | 100% | |
| All three years | 0-64 | 51 (0%) | 10(0%) | 268 (8%) | 176(9%) | 95% | 97% |
| 65+ | 361 (2%) | 463 (2%) | 930 (6%) | 1220 (7%) | 93% | 93% | |
*
Australia, France and Italy were not used for validation analysis because they had shorter time series. Hence leaving out the last three years of data would leave a time series that was too short for estimating model parameters.