Well, let's be careful :-). True, there are numerical methods
(transforms and such ) for a limited set of distribution functions.
But, given huge amounts of computer time, the uniform distribution can
be mapped to any user-defined curve by using simple
weighting/clustering. No argument that this is a poor way to generate a
realistic set of data.
Please ask such non-Mac-specific questions on R-help.
For 'arbitrary', no. But there are lots of helpers, including
Runuran, SuppDists and distrSim, which cover lots of univariate
distribtutions. There are lots of multivariate ones implemented too,
often in packages that make use of MCMC. But you cannot even define
an arbitrary distribution in a unified way.
_______________________________________________
R-SIG-Mac mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/r-sig-mac
