Figure 3
From: Mapping socioeconomic indicators using social media advertising data
Performance of the models across the distribution of Wealth Index. Performance of the models for Philippines (A) and India (B) across the distribution of Wealth Index. For each cluster, the difference between its ranking when ordered according to the DHS wealth index versus when ordered according to predicted Wealth Index was calculated. The data was then split into deciles according to the DHS Wealth Index and the average absolute rank difference computed for each. The y-axis shows the mean absolute rank difference for clusters in that decile as a fraction of the total number of clusters in the dataset. The figures are based on cross-validated predictions from the tree model with Facebook features, log population density and regional indicators