Figure 4

Time-series data showing recorded incidence from PT (black) and AR(1) AIC(i*) modeled incidence (blue) for 2004–2011 for 2 states (North Dakota and New York) with well performing models in the top panels and 2 states (Connecticut and Alabama) with poorly performing models in the bottom panels. The variability between state-model accuracy suggests that GT surveillance approaches cannot be performed uniformly across regions of the U.S.