Within R-sq. in Fixed Effect estimation

[G. van der Laan]

Hi,

From the manual, I understand that the benchmark in computing the R-squared in fixed effects (FE) estimation is the constant only model. I would like to calculate the within R-squared (omitting the explained variance from the FEs).

I thought of first estimating an FE-only model and then estimating the model with the residuals as the dependent, but this makes the procedure two-step as opposed to the one step estimation in the original FE-estimation, which, I think, would alter the results.

Is there any procedure implemented in EViews 6 (or - is there any trick to apply) to compute a within R-squared?

Thanks.

[EViews Glenn]

Just estimate a model with a constant only and then estimate the model with the constant and your other regressors. You'lll have two different SSRs and you're essentially done.