Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Re: xtreg fe cluster and Ftest


From   Kit Baum <baum@bc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: Re: xtreg fe cluster and Ftest
Date   Sat, 26 Apr 2008 06:35:54 -0400

But as Jeff Wooldridge's undergraduate econometrics book "Introductory Econometrics" (now in 4th edition) points out, in many circumstances, F-tests can be 'robustified', or made robust to arbitrary heteroskedasticity. The one we're talking about here is just a test on an OLS model with a bunch of dummy variables. Although xtreg, fe will not give you an F-statistic for joint significance of those variables when robust (actually cluster()) is specified (and now will -areg- with robust), you can always compute it for a standard -robust- estimator if the number of dummies is not too large.


webuse grunfeld, clear
qui tab company, gen(C)
qui reg invest mvalue kstock C1-C9, robust
// this should be the 'robustified' F-test
testparm C1-C9
// for comparison: here is the non-robust F test
xtreg invest mvalue kstock, fe
Note this will not work if you use cluster(company), which is actually the kind of VCE that xtreg, fe robust is employing. But the only difference between robust and cluster(company) is that the latter allows for arbitrary correlation between errors within each cluster.

Kit Baum, Boston College Economics and DIW Berlin
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html


On Apr 26, 2008, at 02:33 , Stas wrote:

F-tests are ratios of variances. When you start talking about
- -robust-, it means you do not think there is a common variance
anymore, so Stata does not provide neither the variances themselves
nor their ratios. A perfectly sensible answer.

You can follow up through the mechanics of the F-test, but what you
will get in the end is a random variable with unknown distribution...
probably a ratio of two complicated quadratic forms in normal
variables, neither of which has a chi-square distribution, to begin
with.
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index