On Sat, 31 Mar 2012, Daniel Bencik wrote:
Following up our discussion, the "best" model for my
problem (under given
circumstances) is AR(7)-GARCH(1,1)-t [...]
The Eviews estimation result is
http://eubie.sweb.cz/gretl_forum/ar7.PNG
Note that you get very similar results using gretl + gig, as
per Jack's suggestion:
<hansl>
open allRng.gdt
include gig.gfn
moo = gig_setup(allRng,1,const,null,7)
gig_set_dist(&moo,1)
gig_estimate(&moo)
</hansl>
However, it appears that the gretl data you supplied are not
the same as those used by Eviews: If I'm reading its output
right, Eviews is reporting 999 observations, of which it ends
up using 992. The gretl data file allRng.gdt has a series
length of 992, with the first 21 observations missing, so that
gig ends up using 971 observations.
If that's right then obviously we won't expect identical
results.
Allin Cottrell