[Carlo Lazzaro]

Lenny:
sorry, but being a homework, my support stops here (https://www.statalist.org/forums/help#adviceextras).
That said, as per your last post, you should probably benefit from studying any decent textbook on panel data econometrics.

[Lenny Tran]

Lenny:
sorry, but being a homework, my support stops here (https://www.statalist.org/forums/help#adviceextras).
That said, as per your last post, you should probably benefit from studying any decent textbook on panel data econometrics.

its not actually homework, it is a research but of course it is "graded" , i dont know how to describe it for you
But thanks anyway, you helped me a lot !

[Iuri Gavronski]

Carlo Lazzaro ,

Out of curiosity, I've seen you recommending -xtregar- in several occasions in this forum. But -xtregar- is not able to deal with heteroskedasticity, am I right? If so, aren't there too many datasets where -xtregar- would be useful. Would you mind explaining why does -xtregar- does not have a vce(r) option?

Best regards,

Iuri.

[Carlo Lazzaro]

Iuri:
-xtregar- outperforms -xtreg- for T>N panel datasets.
It's true that cluster-robust (as well as bootstrap) standard errors are not available for -xtregar- (whereas they are for -xtgls-).
I'm under the impression that -xtregar- focuses on modeling AR(1) nuisance because it can more relevant than heteroskedasticity in T>N panel datasets.
They may also be technical reasons (that I don't know) that make taking heteroskedasticity into account unfeasible under -xtregar-.