Figure 8.

Posterior predictive distribution for the epidemic model using (a) short-time data; (b) long-time data; and (c) short-time data where observations are also made of the number of exposed individuals. In (a,c), the dashed line indicates the last observation point used for inference. The first 3000 samples from each pilot chain is discarded as burn-in. We resample 10 000 parameter combinations (with replacement) and solve the SDE model to estimate posterior predictive intervals (PIs). Shown are 50% (darker) and 95% (lighter) prediction intervals computed from the quantiles of the posterior predictive distribution.