I am currently working with some panel data for a paper. The point of my model is reject or prove 9 different hypotheses.
I have an unbalanced panel with 548 observations over four time periods and 145 different entities. I am trying to decide between pursuing a fixed effects or random effects model.
The Hausman test result indicates that I should use a fixed effects models. I have two problems with this:
1. The related literature typically uses a pooled OLS or random effects approach.
2. The adj. R squared in my fixed effects model is negative, whereas it is 9% (ie. similar to the related literature) in my random effects model.
In other words, I would prefer to use the random effects model but the Hausman test results tells me to do otherwise.
Does anyone have any thoughts on whether using the Hausman test is appropriate?