Fig. 4.

Epidemic growth rate estimates from mechanistic model fits to population-level C_t distributions across the first wave of COVID-19 in three Madagascar regions. (A.) Comparison of COVID-19 epidemic growth rates from March-September 2020, estimated from IPM (blue) and publicly reported (gray) case count data using EpiNow2 (Abbott et al., 2020a) with estimates derived from Gaussian process (GP; red) mechanistic model fit to the time series of C_t distributions (Fig. 3A). Median growth rates are shown as solid lines, with 50% quantile on case-based estimates and 95% quantile of the posterior distributions from C_t-based estimates in corresponding sheer shading. (B.) Growth rate estimates from individual SEIR C_t-model fits to each C_t-distribution shown in Fig. 3A; median growth rates are given as horizontal dashes, with the 95%, 70%, 50%, and 20% of the posterior distribution indicated by progressively darker coloring. Estimates > 0 (indicating growing epidemics) are depicted in gold and < 0 (indicating declining epidemics) in purples. (C.) Raw case count data from the time series (dark = IPM data; light = publicly reported data) is shown for reference. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)