The Performance of Genetic Matching to Reduce Selection Bias in Observational Studies

• In Event: Statistics in the Field: Innovative Applications and Investigations

Abstract

In observational studies, pre-existing differences in the distribution of covariates between treated and untreated groups lead to biased treatment effect estimates if these covariates are also related to the outcome. Genetic matching is a method to remove selection bias in observational studies by matching treated and untreated observations with respect to any number of covariates (Sekhon & Mebane, 1998). The purpose of this study is to evaluate Genetic Matching by manipulating sample size, ratio of treated to total sample size, the magnitude of the relationship between covariates and treatment assignment, and the magnitude of relationship between covariates and the outcome . Four implementations of genetic matching are compared against optimal full propensity score matching under the manipulated conditions.

Authors

• Seyfullah Tingir, University of Florida

• Walter L. Leite, University of Florida