Improved estimation of time-varying reproduction numbers at low case incidence and between epidemic waves
Fig 6
H1N1 influenza transmission in Baltimore (1918).
We compare APEestim (top), EpiEstim with recommended weekly window (middle) (both with Gam(1, 2) prior distribution) and EpiFilter (with m = 2000, η = 0.1, Rmin = 0.01 and Rmax = 10) on the H1N1 influenza dataset from [42]. We use a 5-day moving average filter, as in [42], to ameliorate known sampling biases. Estimates of reproduction numbers, Rs, and corresponding 95% equal tailed credible intervals are in red. One-step-ahead predictions of incidence, Is, (with 95% credible intervals) are in blue with the actual incidence in black. We find that EpiFilter combines the benefits of APEestim and EpiEstim, achieving both good estimates and predictions.