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 at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: Fixed Effects GLS


From   "Andrea Morescalchi" <morescalchi@ec.unipi.it>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Fixed Effects GLS
Date   Tue, 12 Apr 2011 14:27:19 +0200

Thank you vey much Mark. I will look into this interpretation of FE, which I didn't know.

But, regardless of the definition, I still do not understand how to tell Stata to estimate what in the 10 Chapter of Wooldridge (2002) is called FEGLS. Basically, I would like to estimate in Stata a FE where the (T-1)*(T-1) covariance matrix of the error component is not an identity matrix (multiplicated for sigma) but it is estimated in a first stage.

Thanks
Andrea

On Tue, 12 Apr 2011 11:28:47 +0100
 "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk> wrote:
Andrea,

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Andrea Morescalchi
Sent: Tuesday, April 12, 2011 9:57 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Fixed Effects GLS

Dear Ada and Maysa,
I was trying too to estimate the FEGLS as presented in Wooldridge (2002), but with only one error component.

Anyway, I cannot understand properly the first line of Ada's response: "Fixed Effects model is a kind of GLS model. Use the basic FE commands and you will be fine." I red carefully the Chapter 10 in Wooldridge, but I cannot see among the - xtreg, FE - options available which is that for doing FEGLS. In particular, the options available for vce are: conventional, robust, cluster, bootstrap, or jackknife. FEGLS implies a particular structure for the (T-1)*(T-1) covariance matrix for the errors, i.e. the matrix is estimated after a standard FE in the first step where this matrix is an identity matrix multiplicated for the T-constant variance parameter. Among FE Stata options the most proper to do FEGLS would seem: "vce(conventional), the default, uses the conventionally derived variance estimator for generalized least squares regression". This line is a bit strange to me since it seems to imply that the standard FE Stata regression is actually a FEGLS!

It is indeed, and not just in Stata. The FE estimator is, by
definition, a GLS estimator.

I am sure this is discussed in Jeff Wooldridge's book somewhere (there's a 2010 edition out, by the way), but I don't have it handy and can't direct you to the right chapter/page. Another place you might look is Arellano's 2003 book, "Panel Data Econometrics". In chapter 2 (I think) he shows that the FE estimator is the GLS estimator obtained after
applying the first-differencing transformation.

--Mark

I computed the FE and the FEGLS separately in Matlab and it turns out that the "conventional" Stata FE is identical to that with error v-c matrix equal to an identity matrix (multiplicated for sigma_u). But which is the option which tells Stata to estimate the v-c matrix in the first step?

Thanks in advance
Andrea

On Tue, 12 Apr 2011 00:09:11 +0000
  "Ada Ma" <heu034@googlemail.com> wrote:
>Fixed Effects model is a kind of GLS model. Use the basic >FE commands and you will be fine. > > There are user written commands for two way FEs but I >can't remember the commands' names, sorry! There are two >of them and they r both featured in the Stata Journal, >one of the authors for one command is Upward. HTH. > > > Sent using BlackBerry(r) > > -----Original Message-----
>From: May Ster <mayfrank@gmail.com>
> Sender: owner-statalist@hsphsun2.harvard.edu
> Date: Mon, 11 Apr 2011 23:10:06 > To: <statalist@hsphsun2.harvard.edu> > Reply-To: statalist@hsphsun2.harvard.eduSubject: st: >Fixed Effects GLS > > Dear Statalist, > > I tried reading Woolridge (2002), Chapter 10 after he >mentioned > choosing between Random effects model and Fixed effects >model, he does
> mentioned 'Fixed Effects GLS".
> > I'm not sure how does this apply to STATA command if i >want to use
> 'Fixed Effects GLS".
> > If my model is a two-way fixed effects model, Basically, >is the 'Fixed > Effects GLS" can be written like this (i've checked for >serial
> correlation in error terms) ;
> > xi: xtgls var1 var2 var3 i.year , panel(hetero) >corr(psar1) > > > Please help. > > Thank you very much in advance > > Maysa
> *
> *   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/
> > *
> *   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/

*
*   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/



--
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


*
*   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/

*
*   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