Re: [Gretl-users] AR-GARCH-t ML

On Thu, 29 Mar 2012, Daniel Bencik wrote: Out of curiosity: is there any special reason why your specification has to be an arma-garch model with t innovations? From a quick look at the data, it would seem there are better alternatives. For example, your data exhibit quite a few features that are typical of long-memory processes. Just as an example of what I have in mind: <hansl> include gig.gfn open http://eubie.sweb.cz/Allin/allRng.gdt # remove long memory y = fracdiff(allRng, 0.5) # set up an AR(1)-GJR model model = gig_setup(y, 3, const, null, 1) # with skewed-t innovations gig_set_dist(&model, 3) # estimate gig_estimate(&model) </hansl> Riccardo (Jack) Lucchetti Dipartimento di Economia Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti