I Panel study, multiple linear regression, assumptions

[monsmatglad]

Hey.

I am doing a project where I am studying a set of companies over a 7-year period. I am doing a multiple linear regression analysis either with fixed or random effects (so, it's a panel study). What I am wondering is if the general assumptions/requirements apply when using the fixed/random effects technique, so that I should test for them to ensure they are fulfilled?

The assumptions I am referring to are:

- The variables are normally distributed
- The relation between the independent variables and the dependent variables are linear
- homoscedasticity
- independent and normally distributed residuals

(I plan to use a Hausman's test to decide on whether to use the fixed or random effects model)

Thanks in advance
Mons

 

[MarneMath]

Yes and you should also test for correlation since you have a time index in your model. (While that may be assumed by iid portion I find that not many people actually test for it).