Re: st: Question about pstest (after running psmatch2)

[Farasat Bokhari <farasat.bokhari@yahoo.com>]

thanks, but that does not seem to resove the problem:  I used iweight instead of fweight and it produced identical results.  further, i did a list of weights and treatement variables and the values seemed odd.  based on the print out, i created a new weigth variable "mywt" and then tried the regress statement as follows:


---- code ----

gen mywt = _weight
replace mywt = 1 if _treated  == 1 & _support == 1


regress x1 treatmentvar
regress x1 treatmentvar if _support == 1 [fw=mywt]

pstest x1


---- end code ----


now the results from the two regress statements match the output from pstest.  this is some what odd since i thought the original _weight variable is already 1 for treateded cases but inspection showed that it could have a value greater than 1.  now i am not sure if should use the built in _weight variable created after ps2match command.

any ideas??

thanks in advance,
farasat




----- Original Message -----
From: Anca Cotet <acotet@yahoo.com>
To: "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Cc: 
Sent: Wednesday, August 31, 2011 8:04 PM
Subject: Re: st: Question about pstest (after running psmatch2)

you use fweight and pstest uses iweights.




________________________________
From: Farasat Bokhari <farasat.bokhari@yahoo.com>
To: "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Cc: "farasat.bokhari@yahoo.com" <farasat.bokhari@yahoo.com>
Sent: Wednesday, August 31, 2011 8:48 AM
Subject: st: Question about pstest (after running psmatch2)


I have run the following two codes and have a question
about the second one:

psmatch2 treatmentvar x1 x2 x3 x4, out(outcomevar) logit
common
pstest x1 x2 x3 x4

The help file for the pstest states that: 

"For each variable in varlist it calculates (a)
t-tests for equality of means in the treated and non-treated groups, both
before and after matching. T-tests are based on a regression of the variable on
a treatment indicator.  Before matching
this is an unweighted regression on the whole sample, after matching the
regression is weighted using the matching weight variable _weight and based on
the on-support sample"

When i try to do this manually, i can replicate the t-value
result for the unmatched case, however, i get a very different t-value for the
matched case.  I did it as follows:

regress x1 treatmentvar
regress x1 treatmentvar if _support == 1 [fw=_weight]

Cant figure out why the t-value on treatmentvar for the
second regression above does not match the output from pstest. 

Any help/insight would be greatly appreciated. 

*
*   For searches and help try:
*  http://www.stata.com/help.cgi?search
*  http://www.stata.com/support/statalist/faq
*  http://www.ats.ucla.edu/stat/stata/        ;

*
*   For searches and help try:
*  http://www.stata.com/help.cgi?search
*  http://www.stata.com/support/statalist/faq
*  http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/