Bayesian Hierarchical Models for the Design and Analysis of Studies to Individualize Healthcare
Johns Hopkins Bloomberg School of Public Health
Hierarchical Bayesian Modeling; Longitudinal Data; Latent Variable Modeling; Causal Inference; Bayesian Computing; Epidemiology
I conduct statistical research on hierarchical Bayesian latent variable models motivated by and applied to health decisions made for individuals. Broadly, I am interested in discovering simple latent structure representations in complex biomedical data that can improve inferences and decisions about population or individual health. I have also worked on causal inference methods 1) to evaluate novel treatment rules under special designs like matched-pair cluster randomized design, as these designs are useful for interventions that can only be applied at cluster level; and 2) to facilitate the inference for novel estimands and/or semiparametric models by automating and unifying the derivation of efficient influence functions (EIF) and ensuing estimation. Currently a major focus of my work is on analysis of multiple mixed-type longitudinal measurements with feedbacks in treatment assignments. I am working on hierarchical Bayesian methods to to infer latent trajectories that represent individual disease progressions. These methods have direct applications to childhood pneumonia etiology studies, infectious disease surveillance and intervention and psychological researches on depression.