Bases on the following output of Hausmen test in stata, which effect should i prefer between Fixed Effect or Random Effect, Thanks in Advance
b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg
Test: Ho: difference in coefficients not systematic
chi2(5) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= 4.61
Prob>chi2 = 0.4646
(V_b-V_B is not positive definite)
Nothing to do with any kind of test. You'd rather ask yourself if you are ready to consider the effect categories as a sample from a population on which you want to make inference(s).
To answer the question on how to interpret the result of the hausman test:
0.4646 is much higher than any usual statistical significance level. Thus, the coefficients of the random effects model are too far away from those of the fixed effects model. It is, therefore, more likely for the fixed effects coefficients to be correct than the random effects coefficients. This is so, because you need less restrictive assumptions for the fixed effects (allow correlation between the individual term and the coefficients),
If the p-value would be below 0.05, one could argue in case of the random effects. This, however, should always be in part with a theory. Further, in your regression, it will be likely that you want to use robust or clustered standard errors. In this case, the hausman test does no work anymore and you should rely on a theory then anyway...
thank you very much Mr Renaud Lancelot and Mr Jakob Schwerter for your valuable suggestions.
xtreg (variables), fe ewstimates store fe xtreg (variables), re estimates store re
hausman fe re
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation