There has been a growing interest in using observational studies to estimate treatment effects on outcomes where treatment selection is often influenced by covariates. Recently, propensity score matching (PSM) method has increasingly being used to reduce bias in estimated treatment effect for observational studies. Greedy Matching (GM), one of the PSM methods, is widely preferred in many studies because of the calculation simplicity of the method. However, GM is still open to be evaluated in terms of bias reduction and classification performances. For this purpose, data including cigarette usage of 17242 individuals in Turkey were used for the comparison of nearest neighbor, caliper, stratification, Mahalanobis metric, and combined propensity score and Mahalanobis metric matching methods in terms of average standardized bias, bias reduction, and accuracy rate. The stratification-matching method should be preferred for not only low standardized bias and high bias reduction, but also high accuracy rate.
Primary Language | English |
---|---|
Subjects | Statistics |
Journal Section | Natural Sciences |
Authors | |
Publication Date | June 25, 2020 |
Submission Date | January 20, 2020 |
Acceptance Date | May 28, 2020 |
Published in Issue | Year 2020Volume: 41 Issue: 2 |