p score via t effects and balancing proprerty
Hello everybody,
I am a master student at the university of Siena (Italy) and I am trying to estimate the effects of a given university program over a sample of Italian graduates.
I was suggested to estimate a propensity score and then match individuals over the results.
I did it through the Becker and Ichino pscore command and I obtained a specification which satisfies the balancing property.
Being aware of unreliable std errors, I ran the STATA built it command t effects psmatch to perfom matching and obtain the magnitude.
I do not understand whether the t effects commands check for balancing itself or it only confines to respect the overlap assumption.
My question arises as I found out that the t effects command works for different specification which do not satisfy the balancing property according to the pscore command (by becker and Ichino). Is there any goodness of fit measure I can use apart from the teffects overlap command?
Thanks
I am a master student at the university of Siena (Italy) and I am trying to estimate the effects of a given university program over a sample of Italian graduates.
I was suggested to estimate a propensity score and then match individuals over the results.
I did it through the Becker and Ichino pscore command and I obtained a specification which satisfies the balancing property.
pscore treament_variable covariate_list, pscore(pscore) blockid(block) comsup numblo(5) logit level(.005)
teffects psmatch (outcome_variable) (treament_variable covariate_list)
My question arises as I found out that the t effects command works for different specification which do not satisfy the balancing property according to the pscore command (by becker and Ichino). Is there any goodness of fit measure I can use apart from the teffects overlap command?
Thanks
Comment