> As an R user, I can tell R actually does not have any Numerical Analysis
> capabilities "by itself". However, many well-known and well-tested Fortran
> subroutines implementing numerical methods do have R interfaces. This
> includes virtually everything one could expect from a numerical package,
> including quadrature, minimization, interpolation, solution of initial- and
> boundary-value problems, and more. Definitely way more than what a
> statistician would ever need. "The R Journal" has plenty of articles about
> that.
> Apparently, what is missing is Maxima's symbolic computations (although I
> wonder how a statistician could really need that as well.)
Well, check David Andrews' web site
(http://www.utstat.utoronto.ca/david/home.html) as well as his book,
co-authored with James Stafford: "Symbolic Computation for Statistical
Inference" OUP, 2000.
> In general, I
> have serious doubts if R is still a "software environment for statistical
> computing and graphics", as R webpage states. it seems to me more or less
> similar to packages like Scilab, Octave, and others, except that the syntax
> is not Matlab compatible.
I would put that differently: you can easily do with R what you would do with
Scilab and Octave BUT YOU CAN DO MUCH MORE if you're interested in statistics
as illustrated, for instance, by: the model formula, the graphical outputs
(plot methods) of fitting functions, the implementation of the grammar
of graphics, etc.
The key advantage I see in Maxima when compared to R ? in addition to
Maxima's built-in symbolic computation capabilities ? is precisely what
motivates Ihaka to rebuilt a statistical engine on top of CLisp. I want
to be able to create, without too much hassle, compiled functions that I
can call from my "top level". The case was in fact very clearly made in
Fateman et al (1995) "Fast Floating-Point Processing in Common Lisp"
(http://www.cs.berkeley.edu/~fateman/papers/lispfloat.ps) and has
recently been reformulated by Didier Vernat (2006) "Beating C in
Scientific Computing Applications"
(http://www.lrde.epita.fr/~didier/research/verna.06.ecoop.pdf). I'm
working with relatively large data sets (the raw data can be seen as
matrices 10 x 10e6 of integers or floating points) and fair amount of my
pre-processing involves "simple" computations requiring loops. Of course
I could do it in C (that's what I've done so far) but I would be very
interested in finding a more "uniform" solution. And, as a side but
important point, if C is a great language, Common Lisp is ? with due
respect to the late Dennis Ritchie ? much more fun to program with.
Christophe
--
Most people are not natural-born statisticians. Left to our own
devices we are not very good at picking out patterns from a sea of
noisy data. To put it another way, we are all too good at picking out
non-existent patterns that happen to suit our purposes.
Bradley Efron & Robert Tibshirani (1993) An Introduction to the Bootstrap
--
Christophe Pouzat
Laboratoire de physiologie c?r?brale
CNRS UMR 8118
UFR biom?dicale de l'Universit? Paris-Descartes
45, rue des Saints-P?res
75006 PARIS
France
tel: +33 (0)1 42 86 38 28
fax: +33 (0)1 42 86 38 30
mobile: +33 (0)6 62 94 10 34
web: http://www.biomedicale.univ-paris5.fr/physcerv/C_Pouzat.html