RE: re:st: pscore question

["Millimet, Daniel" <millimet@mail.smu.edu>]

I'm not sure how useful it is, but you might look at 

Imai, K. and D.A. van Dyk (2004), "Causal Inference With General Treatment Regimes: Generalizing the Propensity Score," Journal of the American Statistical Association, 99, 854-866

Their estimator (if memory serves) is based on first stratifying the data based on the propensity score, and then fitting any appropriate parametric model in each strata (say, an ordered probit/logit) and then obtaining the overall average treatment effect parameters as an appropriately weighted average across strata.

****************************************************
Daniel L. Millimet, Professor
Department of Economics
Box 0496
SMU
Dallas, TX 75275-0496
phone: 214.768.3269
fax: 214.768.1821
web: http://faculty.smu.edu/millimet
****************************************************

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Dan Kimmel
Sent: Wednesday, February 16, 2011 9:02 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: re:st: pscore question

Dear Statalisters,

Thanks for your suggestions re: my pscore question.  I am now
encountering another problem, which was (indeed) my reason for not
simply using psmatch2 in the first place: does anyone know if there is
a way to use psmatch2 (or some other propensity score module) to model
a nominal- or ordinal-scale outcome?  I am attempting to predict the
effect of exposure to violence in school on students' health outcomes,
where health is measured with a simple 5-category subjective response.
 Thanks,

Dan
--
Daniel M. Kimmel
Department of Sociology
University of Chicago
917.696.2597


On Thu, Feb 10, 2011 at 4:05 PM, Alistair Windsor
<alistair.windsor@gmail.com> wrote:
>
> Propensity score reweighting is a possibility but often it does not adequately correct for the covariate imbalance. See for example "Propensity Score Analysis: Statistical Methods and Applications" by Guo and Fraser. Reweighting does allow easy use of regression for further analysis unlike methods like full matching.
>
> Associated to psmatch2 is pstest which looks at covariate imbalance after matching. I also like to run logit/probit regressions on the treatment over the combined treated and pseudo control groups and look for joint vanishing of the non-constant terms using a Wald or likelihood ratio test.
>
> One further option is to look into optmatch for R. This performs optimal matching using a variety of schemes.
>
> The best bet is to try several techniques and see which work best for your purposes.
>
> Yours,
>
> Alistair
>
>
>
> On 2/10/11 1:33 AM, statalist-digest wrote:
>>
>> Date: Wed, 9 Feb 2011 10:23:19 -0500
>> From: "Ariel Linden. DrPH"<ariel.linden@gmail.com>
>> Subject: re:st: pscore question
>>
>> Daniel,
>>
>> If you already have a propensity score estimated, you can use psmatch2" (a
>> user written program by Edwin Leuven and Barbara Sianesi). This program
>> allows you to use an existing propscore and it will conduct the necessary
>> tests of balance on covariates.
>>
>> You also asked about blocking - there is another program that naturally
>> "blocks", but is flexible enough to allow you to generate your own blocks.
>> This program is called "cem" [Coarsened Exact Matching], a user-written
>> program by Matthew Blackwell and Gary King at Harvard.
>>
>> Given what you are trying to do, I would probably prefer to use propensity
>> score-based weighting over matching. Austin Nichols wrote a nice paper on
>> that " Erratum and discussion of propensity-score reweighting" in
>> Stata Journal  8(4):532--539
>>
>> I hope this helps, and good luck!
>>
>> Ariel
>
> *
> *   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/