Figure 9.

The SEIR model fit and 10-day forecast based on the first 17 days of fall wave of the 1918 influenza pandemic in San Francisco using a negative binomial error structure (<dist1>=3). The upper panel provides the empirical distributions of model parameters along with their 95% CIs. The solid red line is the median model fit. The blue dots indicate the data points. The gray lines, which wrap tightly around the median model fit (solid red line), correspond to the mean of the model fits obtained from the parametric bootstrapping with 300 bootstrap realizations. In contrast, the cyan lines indicate the predictive uncertainty around the model fit, which are used to derive the 95% prediction intervals (dashed lines). The vertical dashed line separates the 17-day calibration (left) and the 10-day ahead forecast (right), which performed well.