Propensity score matching

[Angeliki Skoura]

Hi everyone,

I am a new user of stata and I need your advice if possible.

In order to find the significance of propensity score matching I focus on the z-value derived from probit regression?

More specially, i have used the command "pscore $treatment $ylist, pscore(myscore) blockid(myblock) comsup" where the treatment contain the treated and untreated group (0,1) and the ylist contain my independent variable. My output give the following:

Estimation of the propensity score

Iteration 0: log likelihood = -437345.28
Iteration 1: log likelihood = -437345.28

Probit regression Number of obs = 707170
LR chi2(1) = 0.01
Prob > chi2 = 0.9300
Log likelihood = -437345.28 Pseudo R2 = 0.0000

------------------------------------------------------------------------------
publiccomp | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
apopbtdas | 1.74e-06 .0000199 0.09 0.930 -.0000372 .0000407
_cons | -.4981718 .0015596 -319.42 0.000 -.5012286 -.4951151
------------------------------------------------------------------------------

So, can i say from the z-value/ p-value if there is statistical significance? Or do you suggest another way?


Any recommendations and guidance are much appreciated.

Thanks,
Angeliki

[Angeliki Skoura]

Hello everyone,

I have Stata 16, which may not be compatible with PSMATCH2. That's why I use PSCORE (attnd, attnw, attr, atts, attk). Is this correct in your opinion?

Thanks!

[David Radwin]

You may benefit from reading a few articles and/or books on matching theory and practice. For suggestions, see https://www.statalist.org/forums/for...08#post1242208 and help teffects psmatch.

Also, psmatch2 (Leuven and Sianesi, available from SSC) does work with Stata 16. Stata 16 also has built-in matching programs including propensity score matching.

[Angeliki Skoura]

Thank you @David Radwin!!

Best wishes,
Angeliki