Fig. 2

Direct and indirect effects of residential density on walking for transport (a) and recreation (b). Only significant associations (p < 0.05) are reported. Values represent exponentiated regression coefficients. For example, 1.631 in panel a indicates a 63.1% increase in amount of outside-neighbourhood transportation walking followed by a 1 unit increase in frequency of outside-neighbourhood transportation walking. Significant curvilinear associations between pairs of variables are labelled by the figure number depicting them (e.g., Fig. 3a referring to Fig. 3—panel a, representing the association between neighbourhood residential density and entertainment density). Point estimates and confidence intervals of all examined associations are reported in Additional file 1: Tables S1–S7. * p < 0.05; ** p < 0.01; *** p < 0.001