Online Program Home
  My Program

Abstract Details

Activity Number: 561 - JASA Applications and Case Studies Invited Session
Type: Invited
Date/Time: Wednesday, August 2, 2017 : 2:00 PM to 3:50 PM
Sponsor: JASA, Applications and Case Studies
Abstract #322305 View Presentation
Title: Penalized Spline of Propensity Methods for Treatment Comparisons
Author(s): Roderick J Little* and Tingting Zhou and Michael R. Elliott
Companies: University of Michigan and University of Michigan and University of Michigan
Keywords: Causal Inference ; Propensity Score ; Counfounding by Indication ; Bayesian Inference ; Potential Outcomes
Abstract:

Valid inferences for observational studies require controlling for confounders. Standard regression methods fail when time-dependent confounders are mediators of treatment effects and affect future treatment assignments. We propose Penalized Spline of Propensity Methods for Treatment Comparison (PENCOMP), a Bayesian approach to causal inference in this setting that builds on the Penalized Spline of Propensity Prediction method for missing data problems. PENCOMP estimates causal effects by imputing missing potential outcomes for both treatments with regressions that include splines on the estimated propensity of treatment assignment; the balancing property of the propensity yields a form of double robustness, without the need for weighting. Simulations suggest that PENCOMP tends to outperform doubly-robust marginal structural modeling when the relationship between propensity score and outcome is nonlinear or when the weights are highly variable. Our method is applied to evaluate the effects of time dependent interventions for low urine output due to kidney injury on survival.


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

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association