Re: [Gretl-devel] update on gretl functions and lists (long)
On Mon, 28 Jul 2008, Ignacio Diaz-Emparanza wrote:
At the risk of be boring to Allin and specially, Jack, with this
subject....
given that you are talking about some backward incompatible changes for gretl
2.0, please remember also to change the behaviour of the native arima command
in such a way that when we have regressors (X) and differences as in
arima 0 1 1 ; Y X
gretl automatically takes the difference of the X variable.
You're not boring at all, Ignacio. In fact, I've been thinking a bit about
the ARMA issue. See if the following would be acceptable to you: under the
new syntax I have in mind, we would estimate models of the form
A(L) (y_t - x_t \beta) = z_t \gamma + C(L) \epsilon_t
where
A(L) = (1-\phi_1 L-...-\phi_p L^p) (1-\Phi_1 L^s-...-\Phi_P L^(sP)) \Delta^d \Delta_s^D
B(L) = (1-\theta_1 L - ... - \theta_q L^q) (1 - \Theta_1 L^s - ... \Theta_Q L^(sQ))
with the syntax
arima p d q ; P D Q ; y Xlist ; Zlist
Note that the A(L) polynomial contains the difference operators and is
applied to y_t _and_ to x_t. This should cover all the concerns you
raised. By the way, we'd get rid of the unpleasant asymmetry between
conditional ML (least squares) and full ML.
The modifications to the conditional estimator to accomplish this are not
impossible. The MLE estimator may require some more work.
Would this be acceptable? If it is, I'll start working on this by end of
August/start of September (sorry, no, it can't be earlier), so it won't be
in 1.7.6, which should ship sooner than this (session bugs were quite nasty).
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti