Re: [Gretl-users] tests in gretl
On Fri, 12 Dec 2008, Giovanni Madaro wrote:
How can i do these tests in gretl over a panel data regression?
1.. Anderson instrument irrilevance test;
2.. Stock and Yogo weak instruments test;
3.. Hausman test for OLS unbiasedness.
The Hausman test is available in two ways: (a) run pooled OLS then
use the "hausman" command, or (b) estimate the model using the
random effects option and the test appears automatically.
So far as tests for weakness of the instruments are concerned, at
present we only have the F-test which is applicable for the case
of a single endogenous regressor (as discussed by Stock and
Watson). It would be a good thing to add a more versatile test,
and that'll probably happen before long.
Bear in mind that gretl's scripting/matrix language is flexible
enough to program many tests, so long as they don't involve highly
specialized computations.
Allin Cottrell