R devel - Jan 2000 - Feature requests for princomp(.) : Allow cor() specifications

(all in subject).

If I want to do a PC analysis in a situation with missing data,
I may want to have same flexibility as with "cor(.)",
e.g., I may want

      princomp(x, ...,  use.obs = "pairwise.complete")


Actually, I may want even more flexibility.
Currently,
	princomp(.) 
has

    if (cor) 
        cv <- get("cor", envir = .GlobalEnv)(z)
    else cv <- cov(z)
    edc <- eigen(cv)

and If may want it to use a robust covariance or correlation matrix instead,
e.g.,

    princomp(x, cv = cov.mve(x), .....)

Note that for prcomp() which works on the SVD rather than the
correlation/cov. matrix,  similar flexibility may be required.

Opinions?
Volunteers?

(either one is sufficient to answer!)

Martin
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

> From: Martin Maechler <maechler@stat.math.ethz.ch>
> Date: Mon, 31 Jan 2000 16:21:10 +0100 (MET)
> 
> (all in subject).
> 
Having you been looking at your S-PLUS recently?
> If I want to do a PC analysis in a situation with missing data,
> I may want to have same flexibility as with "cor(.)",
> e.g., I may want
> 
>       princomp(x, ...,  use.obs = "pairwise.complete")
Um. That's a way not to get a valid covariance matrix and hence no
valid PCA.
> Actually, I may want even more flexibility.
> Currently,
> 	princomp(.) 
> has
> 
>     if (cor) 
>         cv <- get("cor", envir = .GlobalEnv)(z)
>     else cv <- cov(z)
>     edc <- eigen(cv)
> 
> and If may want it to use a robust covariance or correlation matrix
instead,
> e.g.,
> 
>     princomp(x, cv = cov.mve(x), .....)
as in

princomp(x, data = NULL, covlist = NULL, weights = NULL, scores = T, cor = F, 
        na.action, subset)
?
> Note that for prcomp() which works on the SVD rather than the
> correlation/cov. matrix,  similar flexibility may be required.
> 
> Opinions?
> Volunteers?
> 
> (either one is sufficient to answer!)
> 
It really is easy.  If there were not a feature freeze on I would do
it right now.

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._