Fig. 2

Model validation in the base scenario. In all panels (except panels d and e) the solid line represents the average outputs from 50 model simulations and the dashed lines represent the 95% confidence intervals. In panels a-c and e, dots represent results from national household surveys [51, 52, 84,85,86,87,88,89,90], with error bars representing 95% confidence intervals. In panel f, dots represent data from national surveys of alcohol use [52, 84, 85, 90,91,92]. In panel e, the ‘unadjusted’ estimate of condom use is calculated from the modelled rate of condom use, applying an odds ratio of 2.2, i.e. assuming the odds of self-reported condom use is 2.2 times the true odds of condom use (the model being lower than the survey data to account for social desirability bias)