On Fri, 1 Apr 2022 at 18:11, Sven Schreiber <svetosch@gmx.net> wrote:
Am 01.04.2022 um 15:49 schrieb John C Frain:
> Your main problem here is a lack of data.  A "Breusch-Pagan test is
> nonsignificant (Chi-square(4) = 1.77291, p= 0.777434)" does not reject
> the null of heteroskedasticity.  With your short series, however, the
> test has very low power. Check what you are doing here.

I believe the Breusch-Pagan test that was referred to is the one you get
with "panspec", so a test of the pooled model against panel RE. It would
be less confusing if those guys hadn't invented various tests!

I'm not saying that anything you wrote is wrong.
You are correct.  The mention of heteroskedasticity confused me.  I still believe that the problem is one of lack of data and the resulting low power of the test.

Take Care

Gretl-users mailing list -- gretl-users@gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave@gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/