st: Re: xttest2

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

<>
Nicolas wrote

> I want to be able to test for heteroscedasticity (preferably Breusch-Pagan)
> after running an instrumented regression on an unbalanced panel.
> 
> When I use xttest2 after ivreg2 I get the following error:
> 
> ***********
> Error: too few common observations across panel.
> no observations
> r(2000);
> ***********
> 
> Does anyone now how I can get around this error? Or maybe a command that
> does work.

-xttest2-  requires that SOME observations are present in all panels, as it is computing the correlations across panels of the residual vectors. So

webuse grunfeld
ivreg2 invest (mvalue = kstock time)
xttest2

however if you create imbalance in the panels, so that there are no years in common across all firms, 

drop if year<1945 & company==1
drop if year >1944 & company==2
ivreg2 invest (mvalue = kstock time)
xttest2

the Breusch-Pagan test cannot be performed, as casewise deletion in the matrix of residuals leaves no usable observations, as the error message says. You can get around the error by restricting your analysis to those periods when all panels are observed; -xttest2- will then work fine.

Kit 
author -xttest2- (SSC)


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/