I am using a cross-dominant panel data set. I apply both the error component models (REM and FEM) and the Kmenta approach (heteroskedasticity and autocorrelation). The Hausman test favors REM over FEM. But then the Likelihood Ratio test (and the xttest3 test command) suggests that panels show heteroskedasticity and autocorrelation. Hence Kmenta should be preferred.
However some variables show considerable high WV (more than 15%),while for some others the WV is relatively low (less than 10%). Shouldn't the fact of having more variables with a high WV point me in the direction that the error component models perform better?
A more general question: is the WV anyhow involved in the choice between error component models and Kmenta? Observing my results, I wouldn't say so.
Thank you for your help.
*
* 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/
