Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 135 - Move Non/Semiparametrics Forward in Causal Inference, Missing Data Analysis, and Data Integration
Type: Invited
Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 11:50 AM
Sponsor: Section on Nonparametric Statistics
Abstract #308078
Title: A Bayesian Nonparametric Approach for Estimating the Causal Effect of a Time-Varying/Dynamic Treatment
Author(s): Michael Daniels* and Kumaresh Dhara and Jason Roy
Companies: University of Florida and University of Florida and Rutgers University
Keywords: causal inference; longitudinal data

Enriched Dirichlet process mixture models (EDPMs) provide very flexible conditional distributions that are particularly useful for modeling the distribution of outcomes given confounders for causal inference. We propose an EDPM with G-computation for estimating the causal effect of dynamic treatment regimes. The 'base model' here consists of the product of parsimonius, parametric, conditional distributions typically used in parametric G-computation. Posterior computations are relatively straightforward and involve almost entirely conjugate (within Gibbs sampling) updates. The EDPM approach is explored via simulations and applied to a real dataset.

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

Back to the full JSM 2020 program