Online Program Home
My Program

Abstract Details

Activity Number: 176 - Bayesian Mixture Modeling, Clustering and Unsupervised Learning
Type: Contributed
Date/Time: Monday, July 29, 2019 : 10:30 AM to 12:20 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #305347 Presentation
Title: Mixed Bayesian Additive Regression Trees for Random Effects
Author(s): Charles Spanbauer* and Rodney Sparapani
Companies: Medical College of Wisconsin and Medical College of Wisconsin
Keywords: Bayesian; BART; Personalized Medicine; Clinical Trials; Random Effects

Bayesian Additive Regression Trees (BART) is a Bayesian non-parametric machine learning ensemble model with excellent out of sample predictive capabilities; yet, its use in biostatistical analyses and biomedical research is in its infancy. One barrier is the implicit assumption of independence between observations that is inherent in the original BART model. In contrast, repeated measures data is common in biomedical research and so using BART may not be justified. For instance, in personalized medicine, BART can be applied to clinical trial data with the goal of finding treatment response heterogeneity, or individualized treatment rules, based on patient characteristics. In many clinical trials, however, the outcome is measured at multiple time points, which may violate the independence assumption. Therefore, I am proposing a model that extends BART to account for both between and within group variation. I will demonstrate different priors for the between subject variance including inverse-gamma, uniform, and half-t. An application to individualized treatment for diabetes subjects will be presented.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program