fixed effect standard errors

[mpiuser]

Hi,

I did a simple un-balanced panel regression including a constant and cross-section fixed effect as regressors. I can recover the fixed effects (up to a constant of course) by viewing them.

How can I recover the standard errors for the fixed effects. The covariance matrix only includes the standard error of the constant.

I realize that I can generate dummy variables and include them explicitly in a simple regression, in which case the covariance matrix will include them. But, I have many thousands of cross-sections so this would be incredibly tedious.

Cheers.

[sdstest]

I was also wondering about this. Did you eventually get some advice? Or did you find the solution yourself?

[EViews Glenn]

As you have discovered, EViews treats the fixed effects as nuisance parameters so that standard errors aren't reported. My recollection is that for non instrumental variables/non-weighted cases, it's not that difficult to compute these after the fact, but you're going to have to go to the literature to find the exact formulae.

[EViews Gareth]

You could use dummies, but you may hit memory problems if you've got a huge number of cross-sections.

You can create the dummies easily using @expand(@crossid), or something similar.