Figure 1.
Open defecation predicts stunting, bivariate linear regression.
Note: n = 112 Indian districts; R2 = 34.8%. The size of the circles is proportionate to the population of the districts they represent. The grey shaded area is the 95% confidence set for the regression line.
Figure 2.
Female literacy predicts stunting, bivariate linear regression.
Note: n = 112 Indian districts; R2 = 48.5%. The size of the circles is proportionate to the population of the districts they represent. The grey shaded area is the 95% confidence set for the regression line.
Figure 3.
District average calorie consumption does not predict stunting, bivariate linear regression.
Note: n = 112 Indian districts; R2 = 0.7%. NSS = National Sample Survey. The size of the circles is proportionate to the population of the districts they represent. The grey shaded area is the 95% confidence set for the regression line.
Figure 4.
Dichotomization reduces statistical power: R2, simulations using NFHS-3.
Note: Observations are 1,000 Monte Carlo samples of 20,000 children under 5 drawn from India’s 2005 National Family and Health Survey. PSU = survey primary sampling unit (local area). The legend reports regression dependent variables.
Figure 5.
Dichotomization reduces statistical power: t-statistics, simulations using NFHS-3.
Note: Observations are 1,000 Monte Carlo samples of 20,000 children under 5 drawn from India’s 2005 National Family and Health Survey. PSU = survey primary sampling unit (local area). The legend reports regression dependent variables.