I'm using fixed effect model on panel data, the best model I managed to prepare have significant coefficients both for random and fixed effect, but apparently adjusted r-squared is negative in fixed effect. That is not the case in random effect model (I have positive value of adjusted r-squared). Although Hausman-test suggest that I should use fixed-effect model.
FE: r-squared 0,127, adjusted -0,157 RE: r-squared 0,07, adjusted 0,067
does this model make sense and can it be interpreted? If not, is there any way to improve this model? I have 900 observations