Matching on the Estimated Propensity Score


This page contains Matlab code (mpsc.m) to compute propensity score matching estimates of the average treatment effect along with standard errors that adjust for first step estimation of the propensity score. See "Matching on the Estimated Propensity Score" (Abadie and Imbens, 2009) for more information. The propensity score is estimated by Logit. This is very preliminary. Use at your own risk.

Usage:   [tau,se] = mpsc(Y,W,X,M)

   Input:   Y  (Nx1)  outcome variable    
            W  (Nx1)  binary treatment indicator
            X  (Nxk)  covariates as they enter the logit estimator of the
                      propensity score. First column should be a column of
                      ones. If interactions or high order terms are
                      required, please include them in the matrix X.
            M  (1x1)  number of matches
 

  Output:   tau  (1x1)  estimated average treatment effect
            se   (1x1)  standard error


The file with the code is here.

Updated 08/24/09