Re: [Gretl-users] GRETL versus Stata: Time Series
On Tue, 26 Apr 2011, Charles Koss wrote:
So, I would like to know some papers that compare the time series
capabilities of Stata and GRETL in order to have a better
understanding of the two softwares. Any help is appreciated.
A special issue of the Journal of Statistical Software is in preparation,
in which a number of statistical packages (including stata and gretl) are
compared with respect to their implementation of linear state-space
models, Kalman filtering etc.
On Wed, 27 Apr 2011, Thomas Volscho wrote:
Does anyone have any plans to (or has anyone) implemented the
Sims-Zha
model (their Bayesian approach to VAR?)
Not exactly the same thing, but I have a semi-finished product
implementing SVARs (but all inference is frequentist, nothing Bayesian). A
few people who had expressed interest received some early versions. I plan
to turn it into a proper function package when I have the time.
On Wed, 27 Apr 2011, Summers, Peter wrote:
FWIW, I've toyed with the idea of translating some of my Bayesian
(VAR &
other) code into gretl, but haven't for a couple reasons: 1) I haven't
had the time, and 2) a lot of this (including Sims-Zha) is available in
R and therefore accessible via gretl. So this is potentially
re-inventing the wheel. It does seem, though, that it would be fairly
easy to implement some basic BVAR functionality in gretl, and I'd be
willing to look more closely at that if there's demand (not before this
summer though).
See above. If you're interested in my SVAR package, I can send it to you.
Actually, it would be nice to co-operate on this.
On the subject of re-inventing the wheel: sure, no-one needs a bazillion
buggy implementations of the same functionality, although you can't stop
people from writing their own mp3 player just for fun. But when it comes
to scientific software, having multiple _free_ implementations of a
procedure is, in my view, a good thing. We all know how buggy software can
be, and being able to compare results from independent implementations is
tremendously useful, in my view, both to the practioner and the coder (who
often coincide). Moreover, I don't mean to diss the R community, but often
R packages are of a much worse quality than core R (IMO outstanding); so,
the mere existence of an R package that claims to do something does not by
itself guarantee that it works properly.
On Wed, 27 Apr 2011, Charles Koss wrote:
The last post in this thread generated some questions:
What incentives could be created for individuals that cooperate with
the development of GRETL?
How to align those incentives to promotion (tenure) and salaries of
those individuals that cooperate in this open source community?
Well, speaking for myself, my primary driving factor was fun. I've been
helping Allin for a few years now and it's been a great gig. Of course,
you also get added benefits, but in my case it was primarily for fun and
the satisfaction you get out of doing something that other people find
useful (as opposed to publishing, on a reputable journal, a clever paper
that nobody is ever going to read, let alone find useful for anything).
However, in the future we plan to shift gretl's development model more and
more towards the implementation of new features through function packages
and add-ons, written in gretl's own scripting language rather than in C, a
bit like R and Stata do. If this development model matures, package
writers who do something interesting and non trivial may have a good
publishing opportunity on specilised journals like the Journal of
Statistical Software, Computational Economics, etc.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti