re: st: Re: heteroskedasticity test in panel data

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

<>
David said

I am estimating a random effects model (xtreg re) after having performed a
hausman test (which indicated that I can use both the fixed effects as the
random effects models) I am now testing my model for the assumptions of
autocorrelations and heteroscedasticity. I have already excluded problems
with autocorrelation.  Additionally I have run a lrtest, following the
guidelines in the FAQ and your anwers to the questions of Jing. The outcome
of this test is reported below. My question is, how should I interpret this
result. I assume that it means that my model suffers from
heteroscedasticity? How then should I proceed further? Furthermore I wonder
(as I am a new to econometrics) whether there are any additional assumptions
that I need to test using the xtreg re model(aside from the normal
assumptions for multivariate analysis)? 

...

. lrtest hetero . , df(18)

Likelihood-ratio test                                  LR chi2(18) =    63.33
(Assumption: . nested in hetero)                       Prob > chi2 =    0.0000


In this case (unlike Jing's) the test procedure from the FAQ works fine and delivers a sensible result.
The null hypothesis for this test is homoskedasticity, which you clearly are rejecting. You should at the
very least use the vce(robust) option. As it appears that you only have 18 panels, it may not be wise
to use cluster-robust standard errors (by panel), but you could try vce(cluster panelid) as well to see
what happens. Clustering by panel would allow for arbitrary dependence between errors within-panel.

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/