Re: [S] Must have S-Plus 4 add ons?

Prof Brian Ripley (
Sat, 7 Mar 1998 08:46:35 +0000 (GMT)

Marc R. Feldesman wrote:
> Now, for my question. I've installed the VR2 libraries, which provide some
> extremely useful added (or replacement) functionality for SPlus. Are there
> any other "general purpose" add-on packages that I should have at my
> disposal. My question is deliberately vague at this point. I don't have
> specific needs that are not currently filled by what I see in VR2 and SPlus
> 4, but you all have worked with S & SPlus for much longer than I have and
> are more familiar with weaknesses that might be filled better by 3rd party
> contributed software of which I may be unaware.
> I can (and have) certainly go through the StatLib archives and see
> submissions that "might" be helpful, but it is unclear to me which of these
> supplement earlier versions of S and which may now be fully implemented in
> the later versions.

Only a very little of what is on statlib has been superseded by later
versions of S -- only survival4 and treefix to my knowledge. BUT as you
are using S-PLUS 4 and hence Windows, the task of porting the Unix versions
will be considerable, and you probably want to confine attention to those
techniques which have already been ported, almost all of which are
in /DOS/S/SWin on statlib.

> My work is specifically multivariate analysis of morphometric data. In my
> typical research, I use both linear and quadratic discriminant analysis,
> principal component analysis, principal coordinates analysis, cluster
> analysis (mostly hierarchical but occasionally k-means), multiple logistic
> regression, multiple linear regression, and both Anova and Manova. I've
> recently gotten interested in the various CART (as a generic term, not as a
> specific term) algorithms, and my colleagues and I have begun to apply
> bootstrapping and randomization testing to some of the sparse data sets
> paleoanthropologists usually work with.

I would suggest you look at the Canty/Davison/Hinkley boot library and
RPart for tree-based methods. (BTW, CART should not be used as a generic
term, it is a trademark for a specific method and software.) You probably
are (or ought to be) interested in exploring non-linear methods, of which
there are several examples in the libraries mda, ppr, polyclass and
polyMARS. Fionn Murtagh had an early collection of multivariate
techniques in library multiv, but there are versions of many of these
elsewhere now.

The boot library is comprehensively documented in the Davison/Hinkley book
on bootstrapping (which I highly recommend). Examples of almost all the
things I have mentioned are in the on-line statistics complements to V&R2.

There is a large collection of tools in the Design/Hmisc libraries of
Frank Harrell which have a medical slant.

Brian D. Ripley,        
Professor of Applied Statistics,
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595
This message was distributed by  To unsubscribe
send e-mail to with the BODY of the
message:  unsubscribe s-news