Guillaume, Saul,
From: "Guillaume Frechette" <[email protected]>
To: [email protected]
Subject: Re: st: Robust covariance matrix
Date sent: Mon, 27 Jan 2003 13:50:29 +0000
Send reply to: [email protected]
> Hi Saul.
>
> Saul wrote:
>
> >I have a panel data set with N (=100) groups and T (=9) observations
> >per group. I want to allow for heteroskedasticity across groups (but a
> >constant variance over time (T) within each group) and arbitrary serial
> >correlation within group but equal pattern aross groups (i.e., the same
> >serial correlation model within each group).
> >
> >I do not want to use GLS. I want to use OLS with robust standard errors.
> >
> >How do I get these standard errors n Stata?
> >
> >My understanding is that using reg with options robust and cluster(i)
> >gives me a more flexible model than the one I want.
>
> I believe your understanding is correct. I do not know of anyone who has
> programmed that correction in Stata. However, I know that John Ham, Steven
> Lehrer, and John Kagel have a paper (forthcoming in the J. of Econometrics)
> where they use a correction somewhere in between what you want and Stata's
> robust, but they coded in Fortran. Maybe the paper is explicit enough about
> how it performs the correction that it would be helpful to you. Good luck.
>
> Guillaume
I'm not sure I understand what Saul is after.
You'd like the same serial correlation model to apply to each group, but you
want it to be arbitrary.
... but my understanding of how these robust covariance estimators work leads
me to think that if the covariance estimator is robust to arbitrary serial
correlation (i.e., you use Stata's -cluster- option), then there's no way to
impose the same model across groups. To do that you have to specify the model,
and that means GLS (and hence not robust).
There *is* an estimator that allows arbitrary time-series behaviour but imposes
homoskedasticity in the cross-sectional dimension. This is the covariance
estimator of Kiefer, J. of Econometrics 1980.
Kezdi has a paper that discusses this covariance estimator and others; you
might find the paper useful.
www.econ.lsa.umich.edu/~kezdi/FE-RobustSE-2002-feb.pdf
Hope this helps.
--Mark
Prof. Mark E. Schaffer
Director
Centre for Economic Reform and Transformation
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS UK
44-131-451-3494 direct
44-131-451-3008 fax
44-131-451-3485 CERT administrator
http://www.som.hw.ac.uk/cert
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
