redundant fixed effects- panel data

[v88]

Hi

I have panel data that has both fixed cross sections and a fixed time period, I am now running a likelihood ratio test on the redundant fixed effects and have now got f-stats for:

Cross-section F
Cross-section Chi-square
Period F
Period Chi-square
Cross-Section/Period F
Cross-Section/Period Chi-square

Could anyone please tell me what these mean ie their equations, their hypotheses and which one is the most important in order to justify choosing the model that I have. I have looked in a few books and none of them mention these, including a book based primarily on eveiws. :?

Thanking you in advance
v

[EViews Glenn]

These are just the standard tests for redundant variables (which happen to be the dummies).

'Cross-section' tests the redundancy of the cross-section effects against the full period/cross-section model.

'Period' tests the redundancy of the period effects against the full period/cross-section model.

'Cross-section/Period' tests the redundancy of both effects against he full period/cross-section model.

The F-versions simply look at the standard F-statistic using the difference of the sum-of-squared (possibly weighted) residuals for the unrestricted model and the restricted. The LR/Chi-squared versions are the standard closely related Chi-square statistics which don't include a scaling term. Any econometric textbook should discuss these tests in great detail.

[v88]

thank you for that.
i have another query now, when i run the correlated random effects-hausman test i get the message that 'cross section test variance is invalid, hausman statistic set to zero' i have tried changing the random effects method on the options tab and still get the same message.

what does this message mean and does it mean that i cannot use random effects in my regression model?

Thanks

[EViews Glenn]

http://forums.eviews.com/viewtopic.php? ... test#p9055

[Starter of EViews]

http://forums.eviews.com/viewtopic.php? ... test#p9055

This question that is how we should deal with when it comes with Cross-section test variance is invalid. Hausman statistic set to zero has been asked many times, but I still have no idea how to fix it. Should I just ignore and analyse using RE model or I should use fixed effect instead? You suggest the reason for that could be small observations, which is difficult to change, so would you minding giving specific solution to that please? Thank you very much.

[EViews Glenn]

The answer to this will depend on your preferences: the question is whether you favor robustness (fixed effects) or efficiency (random effects).

[Starter of EViews]

Thank you very much for your reply. It really helps.