|
Activity Number:
|
127
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Social Statistics Section
|
| Abstract - #304806 |
|
Title:
|
Variable Selection for Propensity Models
|
|
Author(s):
|
Bing Yu*+ and Guanglei Hong
|
|
Companies:
|
University of Toronto and University of Toronto
|
|
Address:
|
30 Charles St. West, 1407, Toronto, ON, M4Y 1R5, Canada
|
|
Keywords:
|
bias ; variance ; MSE ; prognostic score
|
|
Abstract:
|
Propensity score-based methods have been increasingly used in causal inferences with non-experimental data. The purpose of this study is to identify an optimal variable selection procedure. Researchers have proposed alternative strategies. A major contrast is between (a) selecting covariates associated with the treatment and (b) selecting covariates associated with the outcome. Other strategies include (c) selecting covariates associated with the treatment for the propensity score along with additional adjustment for covariates associated with the outcome in the response model, and (d) summarizing covariates associated with the outcome in a prognostic score to supplement the propensity score. We evaluate and compare the performances of these variable selection methods-in terms of reducing bias, variance, and mean square error-using simulated data and real data.
|
