Calibration of forecasts from the semi-mechanistic model.

(A) Calibration of predictive model variants (p-value of the Anderson-Darling test of uniformity) as a function of the forecast horizon. Shown in dark red is the best calibrated forecasting model variant (corresponding to the second row of Table 1). Other model variants are shown in light red. (B) Comparison of one-week forecasts of reported weekly incidence generated using the best semi-mechanistic model variant to the subsequently released data. The data are shown as a thick line, and forecasts as dots connected by a thin line. Dark shades of grey indicate the point-wise interquartile range, and lighter shades of grey the point-wise 90% credible interval.