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:Re: st: Question on xtgls


From   Christopher Baum <kit.baum@bc.edu>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   re:Re: st: Question on xtgls
Date   Mon, 12 Dec 2011 21:41:04 -0500

<>
Steve said

I'm not expert in this area, Phil, but the -help- references www.stata-journal.com/sjpdf.html?articlenum=st0084.  There panels get  dummy variables, -i.e. different means.

Steve
sjsamuels@gmail.com


On Dec 9, 2011, at 1:38 PM, Bromiley, Philip wrote:

I'm having a little trouble understanding xtgls.  The documentation (e.g., page 154 of Stata 12 XT manual) seems to make clear that it is not putting in fixed or random effects for panels but rather is just allowing a complex error structure within panels and across panels.

This seems like a different model that what we normally mean by panel estimators.  We normally want to allow the mean error to vary by panel (in the u(i,t) term).

However, it seems like xtgls retains the assumption that the mean error has expectation zero (with the exception of serial correlation), even though it allows the variance and so forth to vary by panel.

What am I missing?



I'm not sure why the reference to Blackwell's SJ article appears. I think Steve is wrong, in that the [xt] xtgls shows that xtgls with certain options reproduces regress. Consider

webuse grunfeld, clear
regress invest mvalue kstock
xtgls invest mvalue kstock, panels(iid) corr(independent) nmk
xtreg invest mvalue kstock, fe
xtgls invest mvalue kstock i.company, panels(iid) corr(independent) nmk

Note that the regress and first xtgls are identical in terms of e(b) and e(V), and the xtreg, fe and second xtgls, with explicit fixed effects, are likewise identical. So xtgls is
not a fixed-effect estimator, in that it does not allow any coefficient to vary over the panel, including the intercept (unless you put in dummies, of course).

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


*
*   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