Re: [Gretl-users] Changing the limits of the ARMA estimation
On Mon, February 13, 2006 17:51, Ignacio Díaz-Emparanza wrote:
I have a multiplicative ARMA model that seems difficult to estimate
as is
actually not converging. I know that usually it means that the specification
of the model is not good, but I think in this case if I augment the limit of
iterations it will soon converge.
Is there a way of doing that? and, Is there a way of changing the tolerance
criterium for ARMA models?
In the CVS version you should get what you need. Basically, you have two new
variables that you can set: bhhh_maxiter (default value = 100) and bhhh_toler
(default = 1.0E-09); their meaning should be obvious.
Please note that these variables affect every routine that uses the internal
BHHH maximizer, that is ARMA and Tobit at present (although Tobit uses a fixed
tolerance for now --- needs some experimenting). Let me know if this solves
your problem.
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona