Whether xtgls is effective?

[Zhang Hui]

Hi, firends:


I applied the -xgls- command on a data set: mus08cigar, which can be downloaded form bc website. After that, I test regression results with xxtest3. It is still significantly reject the null hypothesis of groupwise heteroskedasticity. Is the "xtgls" still effective? Thanks.

[Carlo Lazzaro]

Zhang:
-xtgls- is the way to go whenever you have a T>N panel dataset with a continuous dependent variable, as in the dataset you quoted.
The main issue is modelling the errors structure.
Otherwise, see -xtgee-.

[Zhang Hui]

Hi, Carlo:

Thank you a lot for your reply and help.