Fig. 1
From: Empirical prediction intervals applied to short term mortality forecasts and excess deaths
Data-splitting setup. 22 years of weekly death counts have been split into 8 overlapping cross-validation data series (CV series). Each series features 5 years of training data and 2 years of test data. The empirical forecasting error is estimated from the test set of the calibration series, validated on the test set of the validation series and applied to the test set of the application series