Re: st: Tests for cross sectional dependence when xttest2 doesn't work

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

<>
On Jun 8, 2011, at 8:33 AM, Rado wrote:

> I would like to test for the presence of cross-sectional dependence in my
> unbalanced panel dataset. However when running the xttest2 command, I receove
> the following error message:
> 
> " xttest2
> 
> Correlation matrix of residuals is singular.
> not possible with test
> 
> Is it possible to overcome this limit of the command? Is there another
> alternative to xttest2?

No. The residual correlation (or covariance) matrix is computed from N vectors of length T in the case of a balanced panel. Obviously if N > T those N vectors cannot be linearly independent, and the rank of the matrix produced must be deficient. This is the same constraint that applies to SUR (-sureg-): you cannot have more equations than time periods per equation. In the case of a large-N, small-T panel, -xttest2- cannot be used.

Kit
author xttest2

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/