Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

Re: st: test for clustering in instrumental variables settings


From   Austin Nichols <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: test for clustering in instrumental variables settings
Date   Sun, 28 Feb 2010 12:22:19 -0500

Sergio I Prada <sprada1@umbc.edu> :
http://ideas.repec.org/p/boc/usug07/07.html
shows problems with not using the cluster-robust SE, and problems
using the cluster-robust SE with unbalanced clusters. There, the FE
estimator with CRSE works better, but it might be worse in your case
to include a bunch of dummies for each cluster (given the probit).
The recommendation in the text there is to construct a simulation
using your data, where you generate clustered errors, and see how your
estimator performs.  This is especially true for your case, since it
is nonstandard!

On Sun, Feb 28, 2010 at 8:28 AM, Sergio I Prada <sprada1@umbc.edu> wrote:
> Stas, Mark, Austin
> Thanks for an interesting discussion.
> Austin: my treatment is hospitals designated as Level I according to the
> American College of Surgeons, all other hospitals in the sample are coded
> 0. (The data comes from a study purposefully designed to study those
> hospitals in comparison to similar hospitals)
> Could you provide one or two references on the discussion about whether to
> cluster or not (pros and cons)
>
> Sergio
*
*   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   |   Site index