xtscc for panel data with fixed effects

[Mark Healy]

Deal All,

I am working with a large panel data (27 countries and 100 years). When doing robustness checks (Pesaran's test and Breusch-Pagan LM) I found there is serial correlation and heteroscedasticity.

xtreg $ylist $xlist, fe
xtcsd, pesaran abs
xttest0

I have used the command xtscc to deal with these problems.

xtscc $ylist $xlist, fe

My questions are:
(i) Is xtscc correct for these cases?
(ii) Is there any other check that I need to do? If not, am I now in a position to report my results?


Best regards,

Mark

[Mark Healy]

Additionally, I would like to know how to interpret the Within R-Squared' reported by the regression with Driscoll-Kraay SE.

xtscc results do not report the adjusted R-squared.

[Carlo Lazzaro]

Mark:
with a T>N panel dataset, your first choice should have been -xtregar,fe-.
That said, .xttest0- does not check for heteroskeadsticity (ie, it is totally different from -estat hettest-), but for random effect after -xtreg,re-.
The community-contributed command (as you should have highlighted in your post, for good reasons explained by a FAQ) -xtscc- actually does not return adj-Rsq.
Explanation concerning the within-Rsq returned by -xtscc,fe- can be found in -xtreg- entry.