Hausman test and dummy variables

[odovakar]

I need help with two things that have been partly answered before on the board. First: I have an unbalanced panel and a fixed effects model in log-log form. How do I conduct an "ordinary" Hausman test in order to see if the chosen specification is correct? Is there some kind of command for it? In Stata this seems to be a piece of a cake, where a single command instantly provides you with P values and all, but I haven't been able to figure out how to do the same in Eviews 5.0.

Second, I have approximately the same problem as another guy who wrote here, i.e. I need to include a column of dummy variables (1 and 0) in the panel, but every time I do this I get the "Near singular matrix" error. Is there some way to get around this obstacle?

Thanks in advance!

[EViews Glenn]

For the first question. Just estimate the random effects equivalent, then use the Fixed/Random effects Hausman test view...

As to the second issue, the deep issue relates to what your dummy variables are meant to represent. There is clearly a singularity going on here that is directly related to the fixed effects. If there are variables that do not vary within cross-sections, then they are not identified in the fixed effects framework.

[odovakar]

Thanks. As for the Hausman effect again - just one, stupid question: How do you do that exactly?
Which settings do I need to use in Eviews for the random effects equivalent? In other words: Are there any specific ones that I need in order for the test to work?
After I have done this: Where do you choose the Hausman test view from?

[EViews Glenn]

In the cross-section effects (I'm assuming you have cross-section effects) setting of the estimation dialog, select Random instead of Fixed.

All of the random effects methods (which are on the options) page are valid for Hausman test comparisons.