Re: [Gretl-users] Problem with Kalman Filter
On Tue, 23 Feb 2010, Marcello Bonassoli wrote:
Hi everybody,
I am using Kalman filter to esteem a particular financial model. The problem is that
Gretl cannot complete
the maximum likelihood estimation (see Gretl script below) and report this message:
Sono state usate derivate numeriche
Tolleranza = 1,81899e-012
failed to invert OPG matrix GG'
The problem is that your algorithm converged to a point where the outer
produce of the gradient is singular. Most likely, not the true maximum
(state space models may have local maxima). Anyway, Allin and I have been
working on related issues recently. A few things you can try:
1) use the CVS version of gretl, or the snapshot if you're using windows,
2) use the --verbose switch with mle, to check that the gradient is
actually 0 at the end of mle.
3) experiment with different starting values.
If you could send me your dataset, I could be able to diagnose your
problem more precisely.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti