Fig. 3: Associations between FPD, markets and household diets.

a, Associations between FPD and household dietary diversity by distance to the nearest urban centre. Marginal effects from correlated random-effect models are shown with error bars representing 95% confidence intervals. The model with data from all countries includes 61,916 observations. Full model results and control variables included are shown in Supplementary Table 3. b, FPD and household dietary diversity in subsistence and non-subsistence households. Subsistence households are defined as those that obtain more than 50% of their food consumption value from own production. The sample with data from all countries includes 78,003 observations. c, Associations between FPD and household dietary diversity obtained from own production and market purchases. Coefficient estimates from correlated random-effect models are shown with error bars representing the 95% confidence intervals. The models with data from all countries include 82,550 observations. Full model results and control variables included are shown in Supplementary Tables 4 and 5. FPD was measured in terms of the number of food groups produced.