Re: st: Murphy and Topel correction and cluster adjustment

[Stas Kolenikov <skolenik@gmail.com>]

You would want to read James Hardin's paper on sandwich estimator of
variance in Stata Journal 2 (3) -- see
http://stata-journal.com/article.html?article=st0018. He showed how to
produce an analogue of -robust- calculation for generated regressors
that is asymptotically appropriate (and, to me, somewhat more
straightforward than Murphy-Topel). I imagine that by repeating his
calculations with sums over clusters rather than over observations,
you will get cluster-corrected standard errors for generated
regressors. You can use the bootstrap, too, but you need to make sure
you do it accurately -- see my paper on survey bootstrap a couple of
Stata Journals ago
(http://stata-journal.com/article.html?article=st0187).

On Thu, Jan 27, 2011 at 8:57 PM, Hongping Tan <hptan@uwaterloo.ca> wrote:
> QVF package provides the MTopel option to compute Murphy-Topel estimator of
> variance. Is it possible to adjust both MTopel and cluster at the same time?
> When I use MTopel option, I find the reported t-stat is too high compared to
> the case when I use CLUSTER option.  Besides, are you aware if there is any
> published paper that uses a bootstrap to solve the generated regressor
> problem?
>
> Thanks,
>
> Hongping Tan
> Assistant Professor of Finance
> School of Accounting and Finance
> University of Waterloo
>
> *
> *   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/
>



-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.

*
*   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/