On Thu, December 21, 2006 13:54, Sven Schreiber wrote: > Allin Cottrell schrieb: > >> >> The estimators you mention are good candidates for inclusion at some >> point, but I don't know how soon -- there are lots of things to be done >> and only two of us actually coding on gretl. >> >> Allin. >> > > I am certainly not against including new features in gretl, but given > the scarce resources, maybe Jack and you should think of a policy that > determines when a new feature should be implemented via the great new > function package mechanism (i.e. doable by many others), and when it has > to be coded (in C) in gretl itself (i.e. Allin or Jack have to do it).
I'm personally inclined to say that from now on everything that can reasonably done via a user function should be done that way. I say "reasonably" because our scripting infrastructure is, as of now, not sophisticated enough to support computationally demanding methods. For example, there is now in CVS a pretty decent implementation of Arellano & Bond's estimator for dynamic panels, which would have been impossible to do via a script. But in the end, it's a bit of a chicken-and-egg problem; gretl's scripting capabilities have increased enormously in a year or so exactly because we started using user-level functions for doing things and discovered what was missing during the process. The other day, I wrote a little function to extract long-run coefficients from an ADL model (attached for those interested), which exposed a bug in the matrix code, now fixed in CVS. The more people write functions, the more enhancements and bug fixes are likely to come in the future -> the easier it becomes to rely on user-level functions for doing stuff. > <Shameless plug:> Of course, allowing shell commands in functions would > greatly expand the scope of function packages by using existing > algorithms in R etc. But it sounded like I already got your "okay" on > that... This is already in CVS. Riccardo (Jack) Lucchetti Dipartimento di Economia Facoltà di Economia "G. Fuà" Ancona
lrcoeff.inp
Description: application/gretlscript
lrcoeff_ex.inp
Description: application/gretlscript
