EViews.com

Hello there

I have a question regarding the estimation of the panel OLS in eviews. As far as I managed to understand the estimation of the panel OLS is based on the LSDV estimator. However I am working on a paper and need to disentagle the portion of adjusted r2 that is due to the fixed effects and the portion that is due to the explanatory variables. A probable answer would be to compute the within adj. R2. Since eviews does not provide something like that can I do it using the output of the equation in eviews? Is there any other way to do that? Moreover, I need to compare the adj. r2 of two different samples for the same model. Is there any build in or add-in procedure in eviews to estimate the cramer (1987) z-test (using the within adj. r2)? Thanks in advance and I apologise for the length of my post.

Best

Kouvas

You can estimate the model with only an intercept and the fixed effects. The SSR from your two models should give you all you need.

There's nothing built-in for the test you cite, but it doesn't appear that it should be difficult to compute given the R2 computations.

Thanks a lot for the prompt reply Glenn ! If I got it right you propose to estimate the full model with fixed effects and a model with only fixed effects and save the RSS of these models. What should be the exact formula to calculate the Within adj. R2 after this? Thanks in advance.

Best
Kouvas

Should be 1 - RSS_full/RSS_fe, which gives you the proportion of the residual from the fixed effects model explained by the full specification including regressors.

Should be 1 - RSS_full/RSS_fe, which gives you the proportion of the residual from the fixed effects model explained by the full specification including regressors.

Thanks a lot for the prompt and very helpful reply.

Best

Kouvas

Just a final question :) The 1-(RSS_full/RSS_Fixed) is equivalent to estimating the full model, then estimate the fixed effects only model and take the difference in the adjusted R2 as the within Adj. R2?

Best

Koubas

I never think in terms of R2, only SSRs, but I think that's going to have the wrong denominator.