You can download the R package hivi to implement the model-based bootstrap for estimation of HIV incidence using STARHS data here.
Modeling HIV prevention approaches for MSM in the Americas: This is a large modeling project evaluating the potential impact of various interventions (male circumcision, pre-exposure prophylaxis, and treatment-as-prevention) on the HIV epidemics among men who have sex with men in the Americas. We use a stochastic, network-based simulation that allows for complex dependence on network structures and individual attributes.
Bayesian inference of contact network structure: In this project, we aim to develop methods for integrating multiple individual-level data sources to make inferences about higher-level properties of contact networks such as cluster or degree assortativity that traditionally cannot be extracted from individual information.
Interference/contamination in community randomized trials: This work comprises two related parts. The first is developing methods to quantify cross-community mixing using viral genetic sequence data, particularly accounting for non-random missingness of sequences. The second aim is to develop methods to leverage measurements of cross-community mixing to estimate the attenuation of the treatment effect due to contamination, a version of causal inference in the presence of interference.
Sensitivity analysis in causal inference: This project aims to develop methods and software to perform analyses of sensitivity to unmeasured confounding in observational studies. Current methods focus on linear models for treatment, but extensions to multilevel models, nonparametrics, and weighted estimators are underway.