re: Re: st: Difference-in-Difference and matching using -diff

["Ariel Linden, DrPH" <ariel.linden@gmail.com>]

While -diff- (findit diff) performs d-i-d, you can easily do that yourself
with -regress- and -margins-. Moreover, you can perform the matching (or
weighting) prior to estimating the d-i-d estimator on your own. These are
straightforward activities that allow you to know exactly what is being
done. 

This seems to be the best answer to you, if you don't "trust" the -diff-
command to do it for you.

Personally, I would perform all these analyses manually, to ensure that it
meets my standards. For example, I may want to test several different
"matching" approaches to determine which one is best (according to
predefined criteria, such as standardized differences or density overlap,
retainment of sample size, etc.). 

Ariel

Date: Wed, 19 Jun 2013 19:32:38 +0800
From: James Bernard <jamesstatalist@gmail.com>
Subject: Re: st: Difference-in-Difference and matching using -diff

Sorry for this. I never realized that there is such thing

- -diff comes from SJ:   http://ideas.repec.org/p/pra/mprapa/43943.html

so is -psocre. It is from SJ :
http://www.stata-journal.com/sjpdf.html?articlenum=st0026

Thanks,
James



On Wed, Jun 19, 2013 at 7:02 PM, Nick Cox <njcoxstata@gmail.com> wrote:
> Although you have very recently had a reminder of our practices you
> still persist in not explaining where user-written programs come from.
> -pscore- and -diff- both need to be explained. Suggested forms are
> given at
>
> http://www.stata.com/support/faqs/resources/statalist-faq/#stata
>
> Many people are lax on this, but that is no reason for you to ignore a
> longstanding request.
>
> Nick
> njcoxstata@gmail.com
>
>
> On 19 June 2013 11:38, James Bernard <jamesstatalist@gmail.com> wrote:
>
>> the user-written command -diff seems like a panacea for all the
>> problems one may face in doing propensity score matching and
>> difference-in-difference. It is fairly simple to implement and there
>> are enough options to work with.
>>
>> A good difference-in-difference analysis involves matching of the
>> control and treatment groups, so that they are very similar in every
>> possible way, except receiving the treatment. Matching is often done
>> via propensity score matching (-pscore) etc.
>>
>> I would like to know if any of you have used this command. How
>> reliable is it? Can I truly say that I did propensity score matching
>> and then did difference-in-difference by just running diff and
>> supplying it with control variables.
> *


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