Figure 1
From: Model fit vs. predictive reliability: a case study of the 1978 influenza outbreak

Forward predictions of the DDE model based on the first \(\tau\) time points (training dataset). Shown are the results for the number of susceptible (S), infected (I), confined-to-bed (B), and convalescent (C) individuals, respectively. The model outcomes (non-black lines) are shown alongside the data (solid black lines). Solid non-black lines correspond to the average trajectories, while dashed lines indicate 95% credible intervals. (b–d) The results are based on the top \(M=100\) (out of \(10^5\)) simulations with the best fit in the first \(\tau\) time steps. (a) \(\tau =14\) (entire epidemic); (b) \(\tau =9\) (peaks in curves B and C reached); (c) \(\tau =6\) (only the peak in curve B reached); (d) \(\tau =5\) (neither B nor C peaks reached). For panels (b–d), RMSE values are calculated based on the testing dataset.