A Robust Multiple Imputation Approach to Causal Inference with Confounding by Indication

• Activity Number: 219 - Seeing the World as a Missing Data Problem: Celebrating 40 Years of Multiple Imputation
• Type: Invited
• Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Roderick J Little* and Tingting Zhou and Michael Elliott
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: Causal Inference; Confounding by indication; Multiple Imputation

Abstract

We describe a robust multiple imputation approach for causal inference called Penalized Spline of Propensity Methods for Treatment Comparison (PENCOMP), which builds on the Penalized Spline of Propensity Prediction method for missing data problems. PENCOMP estimates causal effects by imputing missing potential outcomes with flexible spline models, and draws inference based on imputed and observed outcomes. PENCOMP has a double robustness property for causal effects, and by imputing outcomes of treatments not assigned it can handle confounding by indication, an important issue in causal inference. Simulations suggest that it tends to outperform doubly-robust marginal structural modeling when the weights are variable. We apply our method to the Multicenter AIDS Cohort study (MACS) to estimate the effect of antiretroviral treatment on CD4 counts in HIV infections.