PROGRAMS
Peter Filzmoser

Robust joint modeling of mean and dispersion through trimming:
 (N.M. Neykov, P. Filzmoser, and P.N. Neytchev)
R code
and
information

Imputation of missing values for compositional data using classical
and robust methods:
 (K. Hron, M. Templ, and P. Filzmoser)
robCompositions_1.0.tar.gz Rpackage `robCompositions' (tar.gz file for Linux, Mac) for
running the procedures according to the paper
(preprint).

A contextsensitive method for robust model selection with
application to analyzing success factors of communities:
 (A. Alfons, W.E. Baaske, P. Filzmoser, W. Mader, and R. Wieser)
csselect.R Rprogram for robust model selection according to the paper
(preprint). The program
BRLARS.R
is also needed.

Robust factor analysis for compositional data:
 (together with K. Hron, C. Reimann, R.G. Garrett)
CAGeo08Figs.R Rscripts for generating all figures in the paper.
One also needs the programs
pfa1.R
and
factanal.fit.principal1.R .

Outlier detection for compositional data using robust methods:
 (together with K. Hron)
progsOutComp.R Rscripts for generating all figures in the paper.
One also needs the programs
alr.R ,
invalr.R ,
iso_new.R ,
inviso_new.R , and
drawMahal.R .

twoway.rob
 (together with Christophe Croux, Belgium)
Program for the robust additive and multiplicative fit of twoway tables.
The program is written in SPlus, and it needs the functions
prcomp.rob (robust principal component analysis) and
weight.wl1 (calculating the weights for weighted L1regression).
For drawing robust biplots of a "twoway.rob"object, the function
twoway.bip can be used.

factanal.ccr
 (together with C. Croux, G. Pison, P.J. Rousseeuw; Belgium)
Program for
"Fitting Multiplicative Models by Robust Alternating Regression".
The program is written in SPlus, and it needs the functions
prcomp.rob (robust principal component analysis),
weight.wl1 (calculating the weights for weighted L1regression), and
ccrfit (robust fit of twowaytables).
For drawing screeplots, the functions
screeplot.faccr or
screeplot.fanova can be used.

Robust canonical correlation analysis (CCA):
 (together with J. Branco, C. Croux, R. Oliveira)
cc.ssc Splus program for CCA based on a robust covariance matrix
estimation; uses
mesthub (M estimator)
pp.ssc Splus program for robust CCA based on projection pursuit
cancor.rar R program for CCA based on robust alternating regressions
simcc.ssc Splus program for doing simulations