On Tue, 5 Jun 2007, Chris wrote:
is there any way to combine arima with arch? Is this a stupid
question? When
I run the arima stuff I get a fantastic fit to the data, but when I run the
arch test it shows I definitely have an arch process.
At the moment there is no native command that allows you to run a combined
ARMA+ARCH model in gretl. You have 2 alternatives:
1) If your arma model doesn't include a MA part, just add lags of the
dependent variable as regressors: this will give you conditional maximum
likelihood, which is asymptotically equivalent to full ML.
2) If you really need full ML, you can write a function taking the
observed series and the parameters as input and returning the
one-step-ahead prediction errors plus the conditional variances and stick
that into an mle block. You can use the attached script (rough
proof-of-concept) as a template.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti