xtgls fits panel-data linear models by using feasible generalized least squares.  This command allows estimation in the presence of AR(1) autocorrelation within panels and cross-sectional correlation and heteroskedasticity across panels.
这是命令里的说明，应该是不需要考虑固定或者随机问题。

感谢！！！