Re: st: How do I generate a matched sample with psmatch2 ?

[brendan.halpin@ul.ie (Brendan Halpin)]

On Wed, May 23 2012, Frank Spiegel wrote:

> I have already tried the matching procedure with psmatch 2 with the following code:
>
> psmatch2 familyfirm year class assets, neighbor(1) noreplacement
>
> Unfortunately, the matching does not work that good. For example, I have
> a family firm patent from year 2000 but the matched non-family firm
> patent (identified with the new variable _id) is from year 2003,
> although it should be year 2000. The technology classes are also totally
> random.

psmatch2 does propensity score matching, that is, probabilistic
matching. If you have a small number of cells (the product of the number
of categories of each relevant variable) you may be able to do exact
matching. 

If you have too many "dimensions" you'll need an approximate method such
as PS. psmatch2 fits a probit or logit model predicting "treatment"
(i.e. family vs non-family firm) with the supplied list of variables,
and then matches cases with "controls" which have a similar predicted
probability. If your predictive model is good, cases and controls will
be (in an average way) very similar. If you go this route you need more
variables prediting treatment status. Any variable that may have an
effect on both the probability of treatment and the outcome is useful
here. 

Alternatively, explore Gary King's "coarsened exact matching" (-findit
cem- for details). 

Brendan
-- 
Brendan Halpin,   Department of Sociology,   University of Limerick,   Ireland
Tel: w +353-61-213147  f +353-61-202569  h +353-61-338562;  Room F1-009 x 3147
mailto:brendan.halpin@ul.ie    ULSociology on Facebook: http://on.fb.me/fjIK9t
http://teaching.sociology.ul.ie/bhalpin/wordpress         twitter:@ULSociology
*
*   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/