再给你置个顶，看看有没有会的坛友

Unless there is good reason to do so, we recommend that you employ Inverse std.
dev. weights with EViews default scaling, even if it means you must transform your
weight series. The other weight types and scaling methods were introduced in EViews
7, so equations estimated using the alternate settings may not be read by prior versions
of EViews.

嗯，谢谢大家，问题已经解决了。

命令行输入genr e1=resid
data w1
w1=1/e1^2
ls(w=w1) y c x1 x2