This is not usually done, for two reasons. There are good reasons for thinking that error terms for case i will be correlated over time. Omitted variables that have some degree of stability over time will cause such correlations. In many kinds of research (e.g. typical survey research), the individual cases are not interacting with one another. The second reason has to do with degrees of freedom. If you are not imposing any structure on the correlations, at a given time, if you have n cases, there will be n(n-1)/2 correlations among the residuals. You won't have enough degrees of freedom to estimate all of them. Of course, if you want to treat each wave of observation as a fixed effect, you can estimate that. This imposes so much structure on correlations of errors across panels that you won't have identification problems. David Greenberg, Sociology Department, New York University
----- Original Message -----
From: ash_alankar <[email protected]>
Date: Thursday, June 19, 2003 5:57 pm
Subject: Re: st: RE: Fixed Effects Panel Regression with correlation between panels
> Hi Dolores,
>
> Thanks for the info. But when I mean correlation between panels I
> mean that the errors for unit i are correlated with the errors for
> unit j. I am assuming that the errors are independent across time
> i.e. no autocorrelation.
>
> Do you know how to do this in a fixed effects framework? I was
> thinking of using xtgls with a dummy of ones for each different
> unit
> to capture the fixed effects, but was hoping that there might be a
> better way.
>
> Thanks.
> Ash.
> --- In [email protected], D Jimenez Rubio <djr109@y...> wrote:
> > Hi Ash!!
> > Im not sure if it will help you because I don't know if I understood
> > well your query. First of all, what do you mean by correlation
> between
> > panels? Do you mean autocorrelation of errors? If that is your case,
> > that is, you want to test for autocorrelation, then you can use the
> > option xtregar,lbi. The command xtregar assumes theres autocorr
> of
> order
> > 1. and the command lbi afterwards will allow you to test the null
> > hypothesis of no autocorr (it generates a modified version of
> the
> Durbin
> > Watson test for panel data). I don't know if it is of some help,
> but let
> > me know as im doing right now the same as you, ie, im running
> also a
> > panel data with fixed and random effects.
> >
> > Dolores (University of York, UK)
> >
> > -----Original Message-----
> > From: owner-statalist@h...
> > [owner-statalist@h...] On Behalf Of Ash Alankar
> > Sent: 19 June 2003 15:41
> > To: statalist@h...
> > Subject: st: Fixed Effects Panel Regression with correlation between
> > panels
> >
> >
> > Hi,
> >
> > I have a panel data set and would like to run a fixed
> > effect
> > regression (different constants for each panel)
> > accounting for
> > correlation between the error structure of each panel.
> > It seems that
> > xtreg, fe does not allow for this. Is xtgls the
> > appropriate command
> > to use, i.e. does it implicitly estimate a fixed
> > effects model? (I am guessing it does not and simply
> > restricts to constant to be equal for all individuals (panels)).
> Any
> > insights would help.
> > Thanks in advance.
> >
> > Ash.
> >
> >
> > __________________________________
> > Do you Yahoo!?
> > SBC Yahoo! DSL - Now only $29.95 per month! http://sbc.yahoo.com
> > *
> > * 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/
> >
> > *
> > * 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/
>
> *
> * 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/
>
*
* 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/
