Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: psmatch2??


From   Austin Nichols <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: psmatch2??
Date   Mon, 5 Oct 2009 11:36:25 -0400

MAY BAYDOUN <mbaydoun2002@yahoo.com> :
-psmatch2- on SSC finds matches for treatment cases, not control cases
(except insofar as they are matched to treatment cases, or if you
specfy the -ate- option), which you can see directly by looking at the
variables it creates (all named with a leading underscore by default):

use http://pped.org/lalonde
psmatch2 t age educ marr re74 u74, out(re78) n(10)
su _* if t==1
su _* if t==0
su _* if t==0 & _w<.
psmatch2 t age educ marr re74 u74, out(re78) n(1)
su _* if t==1
su _* if t==0
su _* if t==0 & _w<.
psmatch2 t age educ marr re74 u74, out(re78) n(10) ate
su _* if t==1
su _* if t==0
su _* if t==0 & _w<.

and to use this command, the propensity scores should be bounded away
from zero and one (the example above seems to have a lot of propensity
scores too close to zero for comfort). If you want to run a logit
later, you can construct weights from the propensity score _pscore
which is defined for both treatment and control cases; see e.g.
http://www.stata-journal.com/article.html?article=st0136_1 or
http://pped.org/stata

On Fri, Oct 2, 2009 at 2:05 PM, MAY BAYDOUN <mbaydoun2002@yahoo.com> wrote:
> Dear Statalisters,
>
> I was wondering if you could help me figure something out related to psmatch2. So, the idea here is to find a limited number of untreated participants that are similar on propensity score with the treated group, if I am not mistaken. For the nearest neighbor method, if we chose for example 10 or 15, what does that do. Does it give at most 15 nearest neighbors or exactly 15 nearest neighbor per ID? Also, after doing that we get _n1, _nn, _pscore, _support, _weight. Which one of those can be used to run a logistic regression model with say binary outcome y later on  and take into account the PS or is that irrelevant? If we restrict the sample to those who had _weight variable that is nonmissing, would that do it? Thanks a lot for any help you can provide,

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index