Hello, I was registered in the gretl users list but now I want to
contribute with some code I have been working in. The problem is that
the code is in java and is rather long (10 classes) but the package
would be unique since I do not know any other softare that estimates
SARFIMA models in only one step using pseudo M. Likelihood estimation
(other packages uses the 2 step approximation-first estimating d using a
consistent semi-parametric estimation and then the phi's and theta's).
This is excellent. From the paper, I understand that estimation is carried
out via least squares (eqns. 3.1 and 3.2). IMHO, in a perfect world it
would be preferable to use the Doornik-Ooms algorithm for full ML, but,
hey, this is still much better (as you say) than most alternatives!
I am not the developer of the code, a friend of mine (engineer) made
it
following my directions so i do not know how to re-write it in C.
So... is it possible for you to re-write the entire package in C?
I think it is a great opportunity to expand gretl capabilities and
therefore the ammount of potential users.
Java is a (not so close) relative, so a translation may be possible. Send
the code over and we'll see what we can do.
Sorry, in the last mail I forgot to post the link to the paper that
shows how the software was implemented. The name of the paper is:
?Modelos SARFIMA diarios no estacionarios: aplicación al consumo de
energía eléctrica?... Something like "Non stationary daily SARFIMA
models. Application on the uruguayan electricity consumption" This paper
was presented at the Economic lectures in the Uruguayan Central Bank
last August. You can download the paper (in spanish) from:
http://www.bcu.gub.uy/autoriza/peiees/jor/2008/iees03j3250808.pdf
[
www.bcu.gub.uy]
I understand there's an ocean in between, but I'd like you to consider
presenting your paper in Bilbao next May.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti