Re: st: RE: FGLS vs. OLS

[Jordan Silberman <silberman.stata@gmail.com>]

Thanks Dr. Reed. Stata documentation states that the sureg (seemingly
unrelated regression) command uses FGLS. Therefore, it seems to me
that one should be able to use FGLS to estimate a simple model in
which x predicts y with the following command:

sureg (y x)

If sureg uses FGLS, and if the FGLS coefficients are different from
those of OLS, then you'd expect the command above to yield
coefficients that differ from those of a simple OLS regression.
However, when I use a command like "regress y x" to estimate the same
model with OLS, I get the exact same coefficients (standard errors/p
values differ). Why am I getting identical coefficients here, if the 2
commands use 2 different estimators that should yield different
coefficients?

Thanks,
Jordan

On Tue, Jul 16, 2013 at 12:03 PM, Bob Reed <bob.reed@canterbury.ac.nz> wrote:
> Hi Jordan,
>
> OLS and GLS estimators will produce different estimates.  The formulae are different, as you can check by referring to most econometrics textbooks.
>
> W. Robert Reed
> Professor
> Department of Economics and Finance
> University of Canterbury
> Private Bag 4800
> Christchurch
> New Zealand
> Phone: +64-3-3642846
> Fax: +64-3-3642635
> Email: bobreednz@yahoo.com
> Homepage: http://www.econ.canterbury.ac.nz/personal_pages/bob_reed/
>
> Replications Co-Editor, Public Finance Review
> http://www.sagepub.com/journalsProdEditBoards.nav?prodId=Journal200768
>
> Editor, ISRN Economics
> http://www.isrn.com/journals/economics/editors/
>
> ________________________________________
> From: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu] on behalf of Jordan Silberman [silberman.stata@gmail.com]
> Sent: Wednesday, 17 July 2013 3:50 a.m.
> To: statalist@hsphsun2.harvard.edu
> Subject: st: FGLS vs. OLS
>
> Can anyone tell me if it's correct that coefficients computed from an
> OLS regression should be equal to those computed from feasible
> generalized least squares (FGLS) estimation, while standard errors and
> p values should differ across the 2 methods? I'm interested in
> comparing a single linear model across the 2 methods, so there's no
> "seemingly unrelated regression." Thanks, Jordan
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
>
> This email may be confidential and subject to legal privilege, it may
> not reflect the views of the University of Canterbury, and it is not
> guaranteed to be virus free. If you are not an intended recipient,
> please notify the sender immediately and erase all copies of the message
> and any attachments.
>
> Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
> information.
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/