Dear statalists,
I have estimated the average treatment effect on the treated (ATT) of a certain program employing alternative propensity score matching (PSM) techniques [e.g. Nearest Neighbours (NN) , Kernel propensity, Local linear regression, ridge and Kernel mahalanobis]. With this aim I have used the commands psmatch2 (with bootstrapped S.E.), kmatch (for ridge) and teffects psmatch (in alternative to psmatch2 to have Abadie and Imbens' corrected S.E. for NN) and stored the estimates.
What I would like to do is to compute an average of those ATTs with correspondent S.E. to have a synthetic measure of the ATT of the program across the different models employed.
Any suggestion?
I use stata 16.
Thank you in advance!
Best regards,
Luca