[Gretl-users] Re: Breush Pagan test
On Sun, 14 Jun 2020, Sven Schreiber wrote:
Am 14.06.2020 um 17:48 schrieb Allin Cottrell:
> OK, this is more interesting than I thought (and I'd welcome
> comments/criticism from others).
>
> Salvatore's question is prompted by a difference in results between
> gretl's Breusch-Pagan test for heteroskedasticity (via the command
> modtest --breusch-pagan) and the test as implemented in Eviews.
>
> First point: on the face of it there's a difference between the test
> actually proposed by Breusch and Pagan (Econometrica, 1979), which I'll
> call Test1, and that performed in Eviews, which I'll call Test2.
This old thread from the Stata forum might be relevant:
https://www.stata.com/statalist/archive/2006-04/msg00076.html
So maybe power turns around for non-Gaussian scenarios.
Certainly relevant; thanks, Sven. I hadn't twigged that "Test2" is
equivalent to Koenker's robust B-P version. I re-ran my test script
with uniform errors (should probably try some other cases) and
found:
* Under H0 the original B-P test is "under-sized": rejects at much
less than 5 percent frequency using alpha = 0.05. The size of the
robust version is roughly right.
* Under my H1, error=0.2*x*uniform(), the original B-P test still
rejects with much higher frequency than the Koenker variant.
Allin