Fig. 2
From: Is the lack of smartphone data skewing wealth indices in low-income settings?
Kernel-weighted local polynomial plots of log of monthly household expenditures (in CFA francs) for each wealth quintile (top) and education level (bottom) for base and full model using PCA (left) and polychoric PCA (right). Base model refers to wealth index calculated using only information on whether the household owned any cell phone or not, and a full model refers to wealth index calculated with the addition of data on the number of cell phones owned in a household, whether the household owned a smartphone, how many smartphones the household owned, and whether the household purchased mobile data