解决异方差时，用FGLS法修正模型后的问题

Think of a dataset as having n*T observations, where n is the number of panels and T the average number of observations per panel. To use xtgls, T needs to be large. To use regress, vce(cluster), n needs to be large. To consider using both, both need to be large.
The reasons is: xtgls assumes the heteroskedastic is between panel, and it requires large T to estimate the variance-covariance matrix; while this assumption is not true, xtgls becomes inefficient and the reported standard errors will be incorrect.
On the other hand, vce(cluster) is consistent in either case.