Online Program Home
My Program

Abstract Details

Activity Number: 69
Type: Contributed
Date/Time: Sunday, July 31, 2016 : 4:00 PM to 5:50 PM
Sponsor: Biometrics Section
Abstract #320746
Title: Characterizing Causal Treatment Effect Heterogeneity with Conditional Inference Trees
Author(s): Julian Wolfson* and Lauren Erickson
Companies: University of Minnesota and HealthPartners Institute for Education and Research
Keywords: treatment effect heterogeneity ; decision tree ; conditional inference tree ; matching
Abstract:

The main statistical task underlying personalized medicine is determining which population subgroups most benefit from a given treatment. We propose a technique for characterizing treatment effect heterogeneity which combines one-to-one matching with conditional inference trees. "Pseudo-subjects" are created by matching pairs of individuals with opposite treatment assignments but similar covariate values. The (causal) treatment effect and covariates for each pseudo-subject are respectively given by the within-pair difference in the outcome, and a set of "consensus" covariates for the pair. The pseudo-subject data are then used as inputs to a conditional inference tree; splits in the tree define distinct groups of pseudo-subjects with similar causal treatment effects. Splitting is governed by the multiply-adjusted p-values for marginal associations between consensus covariates and causal treatment effects, which are computed using linear mixed effects models to account for the correlation between pseudo-subjects induced by the matching. We illustrate the technique by applying it to data from a randomized trial evaluating the effects of portion size on weight gain and caloric intake.


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

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association