Hello everyone,

I am working on a binary panel data regression.

I did the Hausman test to see if I should use fixed effects or random effects for a logit regression. The test is telling me that the correct one to use is the fixed effects, however when I run the model, i get this result:

2136 groups (6408 obs = 3/4 of the individuals) dropped because of all positive or all negative outcomes and also 3 important variables are omitted because they have no within-group variance, the sex of the individual and the fathers and mothers education.

These variables don't change through the years for any individual and this is why its happening, they don't have within-group variance.

My question is if this problem is enough reason to ignore the Hausman test, change the model and run one with random effects or maybe there is some other test that i could do.

Also notice that with the random effects all these variables are significant.

Thank you so much in advance.