On Tue, 17 May 2011, Bob McCall wrote:
Thanks! That explains it. When I read the Gretl guide it says k
equals the
number of independantly adjusted parameters. I didn't think to include the
variance. It's more natural to think that k would equal the dimension of the
gradient vector.
I notice that R and Stata do it the same way as X-12-ARIMA and
gretl (i.e. they count the variance among the parameters when
computing the information criteria).
Allin Cottrell