2016-01-15

Younathan Abdia, Ph.D. Candidate, Department of Bioinformatics and Biostatistics, University of Louisville

"Propensity scores based methods for estimating average treatment effect and average treatment effect among treated: A comparative review"

Propensity score (PS) based statistical methods, such as matching, regression, stratification, inverse probability weighting (IPW), and doubly robust (DR) estimating equations, have become popular in estimating average treatment effect (ATE) and average treatment effect among treated (ATT) in observational studies. PS is the conditional probability receiving a treatment assignment with given covariates, and PS is usually estimated by logistic regression. However, a misspecification of the PS model may result in biased estimates for ATT and ATE. As an alternative, the generalized boosted method (GBM) has been proposed to estimate the PS. GBM uses regression trees as weak classifiers and captures nonlinear and interactive effects of the covariate. For GBM based PS, only IPW methods for estimating ATT have been investigated. In this article, we provide an overall review of the commonly used PS based methods for estimating ATT and ATE, and examine their performances in estimating ATT and ATE with PS estimated by logistic regression and GBM, respectively. Extensive simulation results indicate that the estimates for ATE and ATT may vary greatly due to different methods. We concluded that (i) stratification and regression may not be suitable for estimating ATT regardless of the estimation of PS; (ii) matching and IPW usually provide reliable estimates of ATT when PS is correctly specified by logistic regression or when PS is estimated using GBM; (iii) regression may not be suitable for estimating ATE regardless of the estimation of PS; (iv) the estimates of ATE based on stratification, IPW, and DR are close to the underlying true value of ATE, when PS is correctly specified by logistic regression or estimated using GBM.

Stay connected TwitterFacebookLinkedInYouTubeInstagram