Different results for propensity score using neighbor matching every time

[Joie_ray]

Hi,

I am using propensity score matching method for a small exercise but have run into a problem as I am getting different results (bold) each time I run it. The difference are quite large at times. I looked in help for seed and psmatch2 also but have not been able to find an answer.

Following is the command I am using first

set seed 123456789
gen u=uniform()
sort u
psmatch2 treat x y z, outcome(out) neighbor(1) com

----------------------------------------------------------------------------------------
Variable Sample | Treated Controls Difference S.E. T-stat
----------------------------+-----------------------------------------------------------
out Unmatched | .103448276 .128586066 -.02513779 .00990444 -2.54
ATT | .103127079 .158349967 -.055222888 .030392251 -1.82
----------------------------+-----------------------------------------------------------


When I rerun the command dropping u another time I get different output,

drop u
set seed 123456789
gen u=uniform()
sort u
psmatch2 treat x y z, outcome(out) neighbor(1) com

---------------------------------------------------------------------------------------
Variable Sample | Treated Controls Difference S.E. T-stat
----------------------------+-----------------------------------------------------------
out Unmatched | .103448276 .128586066 -.02513779 .00990444 -2.54
ATT | .103127079 .157684631 -.054557552 .030612472 -1.78
----------------------------+-----------------------------------------------------------


Would be really grateful if someone could help me explain and correct the above problem. Also I notice that it is the problem only with neighbor matching and not with radius or kernel matching.

Many thanks,
Joie

[Joie_ray]

I found answer to my post in a different discussion which solved my problem.

http://www.statalist.org/forums/foru...atch2-set-seed

Cheers,
Joie