Fe gls

[Sinem Derindere]

Hello,
My data is suitable for FE model but with heteroscedasticity and autocorrelation, I want to conduct FE GLS model, I write the command like this;
.xtgls y x1 x2 i.id, i(id) panels(hetero) force corr(ar1) first, is this true command for FE GLS . I put i.id as a dummy variable since my model is suitable for FE.
Second in this GLS model the distrubiton is not normal, so what can I do
thank you very much for your help

[Carlo Lazzaro]

Sinem:
welcome to this forum.
1) are you dealing with a N>T or T>N panel dataset?
2) what is not normal? Regressand, pedictors residuals?

[Sinem Derindere]

Thank you Dr. Carlo,
1) My panel data set is T>N (28>25)
2)FE GLS model residuals . I commend .predict residulals then jb residuls after conducting FE GLS model [ .xtgls y x1 x2 i.id, i(id) panels(hetero) force corr(ar1) ]
thank you

[Carlo Lazzaro]

Sinem:
why not considering -xtregar, fe-?
Residual normality is a weak requirement that you can safely skip.

[Sinem Derindere]

When I use xtregar the model breaks down, the coefficients become insignificant.
xtgls model that gives the best result as the theory says about my data

Can you confirm me if putting i.id as regressor makes the model FE ? this is what I doubt first.

[Sinem Derindere]

as I read from book , xtgls is a pooled LS method, however when I want to make it as FE model, I put i.id as a dummy regrossor in the model .

[Carlo Lazzaro]

Sinem:
the following thread might be interesting https://www.statalist.org/forums/for...-xtgls-command

[Sinem Derindere]

Thank you Dr. I will read