[R] Robust PCA?

[Martin Maechler]

>>>>> "BertG" == Bert Gunter <gunter.berton at gene.com>
>>>>>     on Thu, 18 Jan 2007 15:28:47 -0800 writes:

    BertG> You seem not to have received a reply.  You can use
    BertG> cov.rob in MASS or cov.Mcd in robustbase or
    BertG> undoubtedly others to obtain a robust covariance
    BertG> matrix and then use that for PCA.

    BertG> Bert Gunter Nonclinical Statistics 

Indeed. Thank you Bert.

BTW, (for the archives) do note that their is a
"R special interest group" (=: R-SIG) on robust statistics,
and mailing list "R-SIG-robust"
(-> https://stat.ethz.ch/mailman/listinfo/r-sig-robust, also for
 archives) with precisely the goal to foster coordinated
programming and porting of robust statistics functionality in R.

Expect to see more on this topic there, within the next few
days.

Martin Maechler, ETH Zurich


    >> -----Original Message----- From:
    >> r-help-bounces at stat.math.ethz.ch
    >> [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf
    >> Of Talbot Katz Sent: Thursday, January 18, 2007 11:44
    >> AM To: r-help at stat.math.ethz.ch Subject: [R] Robust
    >> PCA?

    >> Hi.

    >> I'm checking into robust methods for principal
    >> components analysis.  There seem to be several
    >> floating around.  I'm currently focusing my attention
    >> on a method of Hubert, Rousseeuw, and Vanden Branden
    >> (http://wis.kuleuven.be/stat/Papers/robpca.pdf)
    >> mainly because I'm familiar with other work by
    >> Rousseeuw and Hubert in robust methodologies.  Of
    >> course,

    >> I'd like to obtain code for this method, or another
    >> good robust PCA method, if there's one out there.  I
    >> haven't noticed the existence on CRAN of a package
    >> for robust PCA (the authors of the ROBPCA method do
    >> provide MATLAB code).

    >> -- TMK -- 212-460-5430 home 917-656-5351 cell