Figure 4
Relationships between predicted (x-axis) and observed (y-axis) log incidence based on models fitted to five different mobility-based incidence time series (panels). The coefficient of determination, \(R^{2}\), associated with each best-fit model is indicated in each panel. Values of observed incidence vary across panels due to the effect of different assumptions about mobility used to transform the residence-based time series to mobility-based time series. For example, log incidence under the assumption of no movement never falls below 0, because there were no fractional cases observed in the raw data. Fractional incidence did occur in the other two time series due to each person’s incidence of disease being partitioned across the towns proportional to assumed mobility patterns. Under the ideal free assumption, the diagonal sets of points are a result of incidence on a given day varying across towns only in proportion to their different population sizes