|
Activity Number:
|
156
- Health Policy Statistics Section Student Paper Award
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Health Policy Statistics Section
|
|
Abstract #317344
|
|
|
Title:
|
Improved Matching via Augmented Propensity Score Estimation
|
|
Author(s):
|
Ernesto Ulloa* and Marco Carone and Alex Luedtke
|
|
Companies:
|
Department of Biostatistics, University of Washington and University of Washington and University of Washington
|
|
Keywords:
|
causal inference;
propensity score;
matching
|
|
Abstract:
|
When estimating causal effects of a binary exposure using observational data one needs to balance the distribution of confounders across treatment arms. Nearest neighbor matching with the propensity score effectively removes the effect of confounding allowing estimation of causal effects. Moreover, it has been shown that matching on the estimated propensity score is asymptotically more efficient compared to matching on the known propensity score due to an efficiency gain that arises from estimating the propensity score (Abadie and Imbens 2016). Nevertheless, the efficiency gain is not sufficient for the matching estimator to attain the asymptotic efficiency bound. We propose adding an additional covariate to augment the propensity score model to maximize its efficiency gain. Further, we show that by adding this covariate the proposed matching estimator attains the efficiency gain asymptotically as one increases the number of matches. We show results from a simulation study and apply our proposed method to estimate the effect of circumcision on risk of HIV-1 infection using vaccine efficacy trial data.
|
Authors who are presenting talks have a * after their name.
