Novel Statistical Methods for Confounded and Incomplete Network Data

Project Summary Studies of public health interventions aimed at controlling the spread of infectious diseases such as HIV/AIDS or COVID19 often face important methodological challenges due to pervasive network dependence, confounding and widespread missing data. Each of these complications has separately received considerable attention, however, methods to tackle them when they coexist are currently lacking. We will develop new statistical methodology, specifically causal identification theory and robust estimation theory which we plan to apply to address pressing scientific questions in infectious disease research using data from two randomized trials and two observational studies with data at hand, where both missing data, and network structure occur including the HAALSA South African Study, the Networks, Norms, and HIV/STI Risk Among Youth (NNAHRAY) study, and the Botswana Combination Prevention Study (BCPP) and the Home-based Interventionto Test and Start (HITS) cluster randomized trial. Success in the proposed research will not only allow for robust conclusions to be drawn from data in the above studies, despite the presence of missing data, the potential for confounding bias, and complex social network structure; it will also provide a methodological template for addressing similar questions beyond these four studies as confounded, missing and dependent data are routinely co-occurring complications in Social and Infectious Disease Epidemiology.