R help - May 2003 - The Wrong Choice: Locked in by license restrictions

A colleague pointed me to this article advocating R as a Matlab 
substitute. Here is the link (deliberately on two lines:

http://searchenterpriselinux.techtarget.com/
originalContent/0,289142,sid39_gci902076,00.html

I'm not a Matlab user, but I understand that it provides a nice front 
end to the Linpack collection of numerical linear algebra routines. My 
friend and I wonder if R can really compete in that league. Does anyone 
know if there are published benchmark comparisons between R and other 
packages like Matlab with the focus on numerical stability rather than 
speed?

Cheers,

Murray

-- 
Dr Murray Jorgensen      http://www.stats.waikato.ac.nz/Staff/maj.html
Department of Statistics, University of Waikato, Hamilton, New Zealand
Email: maj at waikato.ac.nz                                Fax 7 838 4155
Phone  +64 7 838 4773 wk    +64 7 849 6486 home    Mobile 021 1395 862

My experience with Matlab users is that they perfer Matlab because of 
the various toolkits that are available (image processing comes to 
mind).  With respect to standard linear algebra my guess is that Octave, 
Matlab, and R are comparable.  A guess, though....

-roger

Murray Jorgensen wrote:
> A colleague pointed me to this article advocating R as a Matlab 
> substitute. Here is the link (deliberately on two lines:
> 
> http://searchenterpriselinux.techtarget.com/
> originalContent/0,289142,sid39_gci902076,00.html
> 
> I'm not a Matlab user, but I understand that it provides a nice front 
> end to the Linpack collection of numerical linear algebra routines. My 
> friend and I wonder if R can really compete in that league. Does anyone 
> know if there are published benchmark comparisons between R and other 
> packages like Matlab with the focus on numerical stability rather than 
> speed?
> 
> Cheers,
> 
> Murray
>

In a message dated 5/27/2003 7:11:00 PM Eastern Standard Time, rpeng at
stat.ucla.edu writes:
> originalContent/0,289142,sid39_gci902076,00.html
I run MATLAB v6.5 Release 13. In my view, the benefit of Matlab over R depends
on your objectives. I am now using R exclusively, except for solving
differential and partial differential equations which R is weak in. If a
comparable suite of DEQ solvers were available for R, then, in my opinion, R
would be superior to MATLAB for many reasons (too numerous to list).

Both use LAPACK. Prof. Bates Matrix package is a useful complement based on
LAPACK. From a computational statistics point of view, MATLAB cannot compare to
R, R is much much better anf the support on r-news, well, there is nothing like
it for MATLAB. So, unless you need to solve DEQs (IVPs and BVPs), PDEs, and now
delay DEQs, use R.

I have tried to find the fortran versions of the MATLAB ODE suite but have not
been successful. Also, looking at the MATLAB code has not been helpful because
the solvers make extensive use of MATLAB built-ins. Don't get me wrong,
MATLAB is an outstanding product...R is "simply the best" (Tina
Turner)

The Serial Fortran Solvers for ODE Initial Value Problems by Alan C. Hindmarsh
in Fortran (http://www.llnl.gov/CASC/odepack/) would be very nice to have in R
for scientific computing.

There are benchmark comparisions of MATLAB vs. S-PLUS in the s-news archives.

Livin La Vida [R]oca (tranlation [R]ockin and [R]ollin)

oscar

In a message dated 5/27/2003 11:18:33 PM Eastern Standard Time, kjetil at
entelnet.bo writes:
> Can you comment on the benefits of odepack versus lsoda?
The benefit of ODEPACK vs. LSODA is mainly that ODEPACK is a collection of
solvers (A. C. Hindmarsh (1983) "ODEPACK: a systematized collection of ODE
solvers"; in Scientific Computing, ed. R. S. Stepleman et al., North
Holland, Amsterdam, pp. 55--64.) whereas LSODA is a solver. Some problems demand
alternate approaches. ODEPACK sports the Petzold DASSL solver which is of proven
utility for large chemical systems. LSODA however is excellent and quite useful.

/oal