Improved estimation of time-varying reproduction numbers at low case incidence and between epidemic waves
Fig 3
Temporal statistics of small or waning epidemics.
We expand on the results from Fig 2D by decomposing the MSE and one-step-ahead PMSE statistics across the 200 simulated trajectories for every scenario in Fig 2. We do not consider the k = 31 EpiEstim example given its poor performance. We present the k = 7 case, which is the generally recommended EpiEstim setting. We observe that EpiFilter significantly improves on MSE throughout the epidemic trajectory (and not only in periods of low incidence) while maintaining comparable prediction accuracies. Coverage statistics for these scenarios, which are given in Fig C of the S1 Appendix, confirm that EpiFilter also consistently contains the true Rs and Is values within its credible intervals.