Re: st: "how close is close?" -nnmatch- and reviewer complains

[Austin Nichols <austinnichols@gmail.com>]

Mirko Moro <mirko.moro@gmail.com>:
-help nnmatch- says in part:

If keep(filename) is selected, each observation of filename.dta will
hold the following variables:
...
 dist:  The estimated distance between observation i and its match j,
based on the varlist_nnmatch values of each and
            the selected weight matrix.

So you can characterize the distribution of distance between matches, etc.

On Wed, Apr 25, 2012 at 5:29 AM, Mirko Moro <mirko.moro@gmail.com> wrote:
> Dear Statalisters,
>
> I have estimated a regression like:
>
> Y = a + b_1T +b_2C + e,
>
> where T takes the value of 1 if the group was treated, 0 otherwise.
> The groups were randomly selected and a probit regression of T on a
> bunch o characteristics show that they are not different.
>
> The referee asked me to include matching estimators as robustness check.
>
> I use the –nnmatch- command in Stata that uses the algorithm presented
> in Abadie et al. And I presented the results using different numbers
> of neighbours (from 1 to 6).
>
> The referee then added:
>
> "On matching: the authors should provide more details on how good the
> matching is (How close is close? Are the covariates balanced after the
> matching? Are all the observations matched?)"
>
> So I guess I need to show him/her something. As -nnmatch- does not see
> to have any diagnostic, should I use some other codes, like
> -psmatch2-? Any suggestions?
>
> Mirko
>
>
> Abadie et al (2004) Implementing matching estimators for average
> treatment effects in Stata. The Stata Journal (2004) 4, Number 3, pp.
> 290–311.

*
*   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/