"positive or non-negative argument" with panel gls estimate

[cao10008]

when I run panel gls( cross-section effect , cross-section weight),
dependt var is toa, independent is mktpwr, crossid is stoc
the message diplayed is "positive or non-negative argument to function expected in computation of group weight(variance)"

I don't know how to deal with it correctly. The wf1 file is attached, help pls!

[EViews Glenn]

You can see what's going on if you run your model without the cross-section weights and compute residuals, you'll see that for cross-section 600603, you have but a single observation. With the fixed effects added for this cross-section, the residual is identically equal to zero so that we are unable to compute cross-section weights.

[cao10008]

You can see what's going on if you run your model without the cross-section weights and compute residuals, you'll see that for cross-section 600603, you have but a single observation. With the fixed effects added for this cross-section, the residual is identically equal to zero so that we are unable to compute cross-section weights.

Thanks, I have checked my data, it is really as what you discribe. And when I exclude this cross-section , the message diappears.

So when that message pops up, does it suggest that observations for some cross- section be too few, and I should check it?
Under other circumstances, will this message pop up?

[EViews Glenn]

In this context it means that there is at least one cross-section with zero variances for the residuals. There are a variety of reasons why this might occur, one of which was present in this case...

[cao10008]

In this context it means that there is at least one cross-section with zero variances for the residuals. There are a variety of reasons why this might occur, one of which was present in this case...

Thanks a lot, Glenn. Your explaination is very helpful.
I wish Eviews could publish an error message book for us in the future, or just as a part of appendix of "command and programm".