Bayesian Credible Subgroup For Excess Zeroes

Date of Award

4-2022

Degree Name

Doctor of Philosophy

Department

Statistics

First Advisor

Duy Ngo, Ph.D.

Second Advisor

Joshua Naranjo, Ph.D.

Third Advisor

Hyun Bin Kang, Ph.D.

Fourth Advisor

Clifton E. Ealy, Ph.D.

Abstract

A goal of subgroup analysis is to identify which subgroups of the population benefit from a given treatment. Traditional and also many recently introduced approaches to subgroup analysis that include statistical hypothesis testing of treatment-covariate interactions, prediction of personalized treatment effects via machine learning tree splitting algorithms, etc. do not explicitly address multiplicity issues. In this article, we extend the two–step Bayesian credible subgroups approach to assessment of heterogeneity in treatment effects and multiplicity-adjusted benefiting subgroup identification for zero–inflated count data, which are often encountered in medical and public health studies. Our approach is to conceptualize a zero–inflated Poisson regression model with Bayesian variable selection to deal with a number of covariates, and subsequent construction of the Bayesian credible subgroups from the posterior distribution of personalized treatment effects to handle the multiplicity issues. Our approach for count endpoints may be useful for various aspects of personalized medicine: from a number of covariates, we can identify types of patients for assessing the benefit of a treatment in a case where the outcome is an efficacy count endpoint, or the risk of exposure in relation to an adverse count outcome. Simulation studies and analyses of real life data examples were carried out to investigate the performance of our proposed method.

Access Setting

Dissertation-Open Access

This document is currently not available here.

Share

COinS