Re: Re: Re: st: RE: FGLS vs. OLS

[Christopher Baum <kit.baum@bc.edu>]

On Jul 18, 2013, at 2:33 AM, Jordan wrote:

> Suppose that you just want to use FGLS to compute a single model--say,
> a linear regression of y on x. In this case, if you use the command
> "sureg (y x)" in Stata, will sureg estimate the model using FGLS? If
> so, is it correct that this single model estimated with FGLS should
> yield the same point estimate (though different standard errors) as
> what you'd get from a simple OLS regression?

It would be meaningless to do so, as the FGLS nature of SUR refers to its formula as a systems estimator. If you're estimating 
a one-equation 'system', it will do nothing else than straight OLS. FGLS on a single equation usually refers to a heteroskedastic
transformation (such as is achieved by Weighted Least Squares and Stata weights) or a transformation dealing with serially
correlated errors, such as is done by Stata's prais command or (for other than AR(1)) Stata's arima command. Those are both
FGLS approaches. But sureg does not do FGLS on a single equation in any meaningful sense.

Kit


Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                             An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
  An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html
                                                                                                   | http://www.crup.com.cn/Item/111779.aspx	


*
*   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/