Dear Statalist,
I have an unbalanced panel data set with 67 firms and 20 years. I have performed several tests, such as the Hausman test, a test for autocorrelation (xtserial), and a test for heteroskedasticity (xttest3) that indicate XTGLS with fixed effects is the correct procedure in my case.
Hence I'm running the following regressions:
xtgls y x1 x2, panels(heteroskedastic) corr(ar1) force
xtgls y x1 x2 x3, panels(heteroskedastic) corr(ar1) force
How can I test in this case if the introduction of the variable x3 improves the goodness of fit as neither the loglikelihood nor the r2 is given?
Thanks for your help,
Thorsten
*
* 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/
