Figure 5.

City-level summary of full-model fits for three cities. Models were fitted using neighborhood-level data but aggregated to the city level for presentation. Observed data (black lines; solid: in-sample, dashed: out-of sample); model predictions (blue lines; solid: in-sample, dashed: out-of sample), 97.5% credible intervals (red shades: in-sample; pink shades: out-of-sample). Insets display the neighborhood-level fits (black points: in-sample data; pink points: out-of-sample data). Best full model from each city is presented. (A) Vitoria: Y _i,t = β ₁ M _i,t-13 + β ₂ Y _i,t-1 + β ₃ M _i,t-1 Y _i,t-1 + β ₄∑_j M _j,t-5 ^0.1 + β ₅∑_j(Y _j,t-1 /d _ij) ^0.5 + β ₆∑_j(M _j,t-1 Y _j,t-1 /d _ij) ^0.1 + log(P _i) + π _i. (B) GV: Y _i,t = β ₁ M _i,t-2 + β ₂ Y _i,t-1 + β ₃ M _i,t-4 Y _i,t-4+ β ₄∑_j M _j,t-1 ² + β ₅∑_j(Y _j,t-1 D _j) ^0.1 + β ₆∑_j(M _j,t-1 Y _j,t-1 D _j) ^0.5 + log(P _i) + π _i. (C) SL: Y _i,t = β ₁ M _i,t-6 + β ₂ Y _i,t-1 + β ₃ M _i,t-1 Y _i,t-1 + β ₄∑_j M _j,t-6 ^0.1 + β ₅∑_j(Y _j,t-1 R _j) ^0.001 + β ₆∑_j(M _j,t-1 Y _j,t-1 R _j) ^0.1 + log(P _i) + π _i.; notation is as in Figure 4.