R-alpha: Mutivariate Analysis

Kurt Hornik Kurt.Hornik@ci.tuwien.ac.at
Thu, 21 Aug 1997 23:02:06 +0200

>>>>> Ross Ihaka writes:

> I have got a little side-tracked (from graphics) and am putting
> together a little multivariate analysis library.  This is just
> intended to be a "core" library rather than anything exhaustive.
> Mainly it is a matter of putting togther code which already exists at
> StatLib.  Here is my present list (only some of which is finished).

>  1. Principal Components
>     prcomp
>  2. Clustering
>     dist, hclust, plclust, subtree, cutree
>     kmeans

>  3. Canonical correlations (is this ever used?)
>     cancor

>  4. Scaling
>     cmdscal, sammon, isoscal

>  5. Graphics
>     (Optimal) profile plots
>     biplots
>     stars etc
>     A bunch of Michael Friendly's stuff converted from SAS

>  6. Discriminant analysis
>     discr (a real one which takes prior probs and returns posterior
>     ones) 

> I would also like to use the object facility so that printing,
> plotting etc is done with generic functions; e.g. plot(prcomp(x))
> should produce a scree plot, coef(prcomp(x)) should deliver the
> loadings - plclust() should also really be plot.hclust().

Yes I know this is a very old mail from Ross ...

I have two questions.

* Is the (already existing) plot.hclust() already a replacement for
plclust()?  (I.e., could we add

	plclust <- function(...) plot.hclust(...)

for compatibility purposes?)

* I'd be very much interested in some of the functions that are not yet
in the current `mva'.  Has there been any work on this one lately?  How
can we help?

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch