Fig. 2
Empirical distribution of random variation in apparent effect sizes for symptomatic seropositivity rate, for symptom rate, and for population size, under simulated random treatment/control assignments. The x-axis range is the same across all three histograms. Consenting population is subject to considerably less variance from randomization than the other outcomes. The histograms are generated by reassigning treatment/control within village pairs randomly, 1 million times, to generate 1 million alternative splits between treatment and control groups. The red line shows the effect size for the actual treatment/control group split. Light blue rectangles contain 95% of the data