|
Session Title: Propensity Score Matching: Further Methodological Development
|
|
Multipaper Session 300 to be held in PRESIDIO A on Thursday, Nov 11, 1:40 PM to 3:10 PM
|
|
Sponsored by the Quantitative Methods: Theory and Design TIG
|
| Chair(s): |
| Ning Rui,
Research for Better Schools, rui@rbs.org
|
| Discussant(s): |
| Frederick Newman,
Florida International University, newmanf@fiu.edu
|
|
On the Bandwidth of Propensity Score Caliper Matching
|
| Presenter(s):
|
| Wei Pan, University of Cincinnati, panwi@ucmail.uc.edu
|
| Abstract:
Caliper matching is one of the most efficient matching techniques in propensity score (PS) analysis. Researchers have been using the fixed caliper bandwidth c = .2 or .25 of the pooled standard deviation of PS, suggested by Cochran and Rubin (1973) and Rosenbaum and Rubin (1985), to address the variation of PS. Unfortunately, the fixed caliper bandwidth can only capture the between-subjects not the within-subjects variation of PS; and on the latter, it is exactly what PS matching operates. The present study proposes a random caliper bandwidth utilizing bootstrap confidence intervals of PS to capture the both between-subjects and within-subjects variations of PS. For subject i, the random caliper bandwidth c_i is determined by the width of the confidence interval of the bootstrap PS for that subject. The random caliper bandwidth is illustrated and discussed with an empirical example.
|
|