[Gretl-users] Re: Breush Pagan test
Am 15.06.2020 um 20:22 schrieb Allin Cottrell
I tried another run of my Monte Carlo, with t(12) errors. Original
B-P
was somewhat oversized (around 0.075 or 0.08 for nominal 0.05) but
reasonably powerful. The Koenker robust variant was correctly sized at
0.05 (as claimed) but had about 1/2 to 5/8 the power against my H1 (with
error multiplied by 0.2*x).
Well if you size-adjust the first test by multiplying with 0.05/0.08 you
also arrive at 5/8 of the raw power. So maybe not so different.
It seems the subsequent literature has kind of skated over
Koenker's
1981 caveat, that the statistic is "not entirely satisfactory" since
"the power of the resulting test may be quite poor except under
idealized Gaussian conditions". I would add, _even_ under Gaussian
conditions.
So what about Wooldridge's F version? It uses the same auxiliary
regression as Koenker and simply takes the ordinary F statistic from
that. Although I fail to see why that would have more power, but who knows.
cheers
sven