[6493] in s-news-athena
Re: S+ vs. Gauss
daemon@ATHENA.MIT.EDU (Dr. Stat)
Thu Feb 23 00:17:41 1995
Date: Wed, 22 Feb 95 23:05 CDT
From: "Dr. Stat" <DRSTAT@polisci.wisc.edu>
To: SWAITJD@dale.cba.ufl.edu, S-NEWS@UTSTAT.TORONTO.edu
X-Vms-To: IN%"SWAITJD@dale.cba.ufl.edu",DRSTAT
Gauss is a programming language, though you can buy add-on modules
that do statistical analysis functions. S+ is a statistical system
that gives easy access to writing extensions. (Or at least that's how
I think of the two.)
I use Gauss when teaching maximum likelihood methods because the
MAXLIK routine is easy to call and the matrix expressions are
"natural" when teaching students. I also use it for my statistical
work when I have to solve non-standard problems.
S+ will maximize a likelihood function supplied by the user quite
nicely. However, it does not have the built in features that Gauss's
MAXLIK does for the estimate of the variance-covariance matrix of the
estimates. (Gauss will provide any of three alternative estimates of
the v-c matrix, including White's heteroskedastic-consistent estimate,
just by setting a switch.) For me, this makes S+ more trouble to use
for this purpose.
S+ gets the clear win as a statistical system. If you really want to
do statistics, and you don't mind having to write the functions to
compute the v-c matrix, then S+ would be the choice. If you want easy
ML (and the new version on MAXLIK, due out in a few weeks looks really
nice with several improvements) then Gauss is nicer.
The right solution is to get both and use which ever makes solving the
problem at hand easiest. (Assuming a lot of money and twice the
learning time.)
Perhaps others with more ML experience in S+ than I have will offer
their views. Also, if there are any ML functions available for S+ I'd
like to hear about them.
(An article comparing S+ and Gauss appeared in a recent issue of The
Political Methodologist. Perhaps the author would care to post it
here. (Are you reading this, Simon?))
Dr. Stat
-------------------------------------------------------------
DrStat@polisci.wisc.edu
Answers Your Questions About Statistics and Political Science
-------------------------------------------------------------
