Fitting dynamic models to epidemic outbreaks with quantified uncertainty: A primer for parameter uncertainty, identifiability, and forecasts

Chowell, Gerardo

The best fit of the GGM model and the corresponding residuals using the first 15 weeks of data of the Ebola epidemic in Sierra Leone is shown in Fig. 2 . Our estimate for the scaling of growth parameter p indicates that the early growth pattern of the epidemic in Sierra Leone followed polynomial growth dynamics (Chowell et al., 2015) . However, we still need to assess parameter uncertainty to construct confidence intervals and diagnose any potential parameter identifiability issues (see Section 9). While the GGM gives a good fit to the data, the EXPM model deviates systematically from the early growth phase, which is evident from the temporal autocorrelation in the set of residuals (Fig. 3) .