|
Activity Number:
|
202
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #309635 |
|
Title:
|
Propensity Score Matching for Causal Inference When Multiply Imputing Missing Covariate Data
|
|
Author(s):
|
Robin Mitra*+ and Jerome P. Reiter
|
|
Companies:
|
Duke University and Duke University
|
|
Address:
|
115 Old Chemistry Building, Durham, NC, 27705,
|
|
Keywords:
|
Propensity Score ; Missing data ; Multiple Imputation ; Observational Study
|
|
Abstract:
|
Propensity Score Matching is used in observational studies to balance the distribution of measured covariates for treated and matched control units. This enables inference about the treatment effect to be based on comparable groups. When units have only partially observed covariates, propensity scores cannot be directly estimated. Multiply Imputing missing values will allow propensity scores to be estimated from the imputed data, however we may be sensitive to the accuracy of the imputation models. In this talk we present some ideas that reduce the dependence of propensity score matching on the choice of imputation models used and can be implemented together with standard Multiple Imputation Software for missing data.
|
