Hi there,
I am running xttobit on a panel (20,000 individuals and 90 time periods) and getting promising results.
However, I am aware that random effects is inconsistent if the random effect is correlated to our explanatory variables (usually tested through a Hausman test).
Would someone be able to point me in the direction of a Hausman test that allows xttobit to be tested? As no fixed effects tobit model exists (due to bias), I have no model to compare the xttobit results to.
If, on the other hand, no Hausman test exists for xttobit, how is it used? I'm under the impression that in order to use any random effects model, this test should always be conducted.
Thanks
David
I am running xttobit on a panel (20,000 individuals and 90 time periods) and getting promising results.
However, I am aware that random effects is inconsistent if the random effect is correlated to our explanatory variables (usually tested through a Hausman test).
Would someone be able to point me in the direction of a Hausman test that allows xttobit to be tested? As no fixed effects tobit model exists (due to bias), I have no model to compare the xttobit results to.
If, on the other hand, no Hausman test exists for xttobit, how is it used? I'm under the impression that in order to use any random effects model, this test should always be conducted.
Thanks
David
