[R] Numerical Recipes in R

[Duncan Murdoch]

On 2/9/2007 5:00 AM, Lorenzo Isella wrote:
> Dear All,
> So far I have mainly used R for data analysis and simple numerics
> (integration of functions, splines etc...).
> However, I have recently been astonished at finding out that many
> things I thought were only achievable with Fortran or C can be done
> e.g. entirely using MatLab.
> When I try asking around if the same could be achieved by R,
> inevitably the answer is that either people do not know R or there is
> so much numerical MatLab code (for instance for solving partial
> differential equations), that there is no point in switching to R.
> Does anyone know if this is really the situation? I am wondering if
> there is anywhere a kind of freely available collection of reliable
> numerical software written in R which is not only geared towards
> statistics and data analysis.

CRAN is the place to look, but I wouldn't expect to find PDE solvers 
there.  (I haven't looked, they may well be there.)  Matlab does a good 
job at those, and most users who want to do that sort of calculation are 
familiar with Matlab, so there isn't much motivation to rewrite those 
things for R.

On the other hand, if you already have Fortran or C code to do these and 
you like it, it is not very hard to write an R interface to it, so you 
can put together your own package and access that code from R.  I think 
this would be a big improvement over standalone Fortran or C (because 
the data handling and graphics functions in R are much easier to use 
than most Fortran or C libraries); whether it would be an improvement 
over Matlab really depends on what you want to do, how good your 
existing code is, and how well you already know Matlab.

Duncan Murdoch