[Michael Lee]

Carlo:
1.So, when I have a long unbalanced panel data (with heteroscadiscity, cross-sectional dependency, and serial correlation or different combination of these problems), which model should I choose?
a) xtgls
b)xtregar
c) xtpcse
d) GMM!

2.and with which options for the command?
I'm not a big fan of-fe-, just hausman test after simple xtreg suggests that.
Thank you in advance for your reply.
Sincerely,

[Carlo Lazzaro]

Michael:
I would try -xtpcse-.

[Michael Lee]

Carlo:
Thank you very much!
And is it needed to add i.firm or i.month to the model?

[Carlo Lazzaro]

Michael:
from the -xtpcse- entry, Stata .pdf manual, the answer is: no.

[Michael Lee]

Thank you very much Carlo.