Fig. 3: Illustrative model shortcomings.

Pitfalls due to collinearity in covariates (left), too much flexibility in mobility (middle), and too much flexibility by including temperature (right). Observed covariates and infection rates from two counties are used to demonstrate these limitations; the same data is used for both overflexibile examples. Observed covariate values (top), estimated time-varying effects (middle), and fitted and observed growth rates (bottom) are shown with median (solid lines) and 95% quantiles (shaded).