UCLA

This dissertation comprises multiple methodological advances in a number of important statistical areas that have clear applications to public health policy and clinical care. I present an overview of the dissertation in Chapter 1. In Chapter 2, I develop a sensitivity analysis framework to extend causal inferences of a binary point treatment from a randomized controlled trial to a target population when a subset of effect-measure modifiers is measured only on trial subjects. In Chapter 3, I develop a Bayesian procedure to extend inferences of the complier average causal effect of a binary point treatment from a randomized controlled trial to a target population. In Chapter 4, I develop a semiparametric method to perform causal inference for a binary point treatment in observational study settings on a recurrent event outcome with respect to a binary point treatment subject to truncation by death and censoring from loss to follow-up, utilizing latent random effects to serve as a collective proxy for potential omitted risk factors and/or effect-measure modifiers of recurrent event incidence and/or death, each of which may also influence treatment selection. Finally, in Chapter 5, I develop of a joint model of univariate longitudinal biomarker and panel count data with an accompanying semiparametric maximum likelihood estimation and inference procedure. The first two projects are centered around the theme of generalizability and transportability, the first three projects are related to causal inference, and the third and fourth projects both deal with the analysis of recurrent event data. Compelling real-data examples are used throughout this dissertation to concretely illustrate the potential real-world impacts that the statistical contributions made by this dissertation can have in informing public health policy and guiding clinical care for patients.