Fig. 1

Overview of emulator development and validation process

Figure 1 depicts the steps used to develop and validate the emulator approximating epidemiological HIV and NCD data from two established simulators. First, the R statistical package is installed. Second, relevant parameters related to the 10-year prevalence of HIV, hypertension, comorbid HIV and hypertension, and depression are abstracted from published simulators to serve as the emulator’s design points, and entered into R. Third, Gaussian processes (GP) approximate the mean and variance of each simulator’s outputs, assuming presence of noise for the emulator’s design points. Fourth, Bayesian posterior predictive analysis is used to infer the credibility of future HIV and NCD prevalence(s) based on the posterior distribution. Fifth, leave-one-out cross validation confirmed the emulator’s predictive accuracy. Lastly, the emulator’s predictions are compared and interpreted in relation to those of the simulator(s)