[R] correspondence analysis and canonical correspondence analysis in R

Jan de Leeuw deleeuw at cuddyvalley.org
Sat Nov 26 05:31:05 CET 2005


http://www.cuddyvalley.org/psychoR/code

now contains version 1.4 of ca.R

There are already many versions of CA and CCA in various R packages,
but this one has some unique features, so maybe it is useful.

The function ca() can do simple CA, but it also
allows for linear restrictions on the row and column scores (they
can be restricted to be in the span of a number of covariates). Thus
the program can also do CCA (and more, because column scores can
be restricted as well).

I adapted the canonical partition of chi-square to restricted
CA, and generalized the notion of Benzecri distances
that are approximated from below. Row-wise and column-wise
partitionings of the inertias are also printed.

The CA program can make row plots, column plots, joint plots,
regression plots, transformation plots, and Benzecri plots.

It allows for four different scalings: x in the centroid of y,
y in the centroid of x, Goodman scaling, and Benzecri scaling.

This still needs some work to make sure everything makes sense
for linearly restricted scores -- and I need to add some plots
specific to CCA.

Updates will be posted to members of

http://www.stat.ucla.edu/mailman/listinfo/albertgifi

The directory

http://www.cuddyvalley.org/psychoR/

also has updated code for logistic unfolding, for
distance association models, for multidimensional
scaling using the smacof algorithm and for multidimensional
scaling using alscal.

==========================================================
Jan de Leeuw, 11667 Steinhoff Rd, Frazier Park, CA 93225, 661-245-1725
.mac: jdeleeuw ++++++  aim: deleeuwjan ++++++ skype: j_deleeuw
homepages: http://www.cuddyvalley.org and http://gifi.stat.ucla.edu
==========================================================
The Good -- and this is surely true --
  is just the Bad that we don't do !          (Wilhelm Busch)




More information about the R-help mailing list