I'd like to announce the availability of a new library for hybrid
hierarchical clustering, "hybridHclust". The library has been
uploaded
to CRAN and is now available.
The library implements a hybrid of top-down and bottom-up hierarchical
clustering. Along the way, the idea of a "mutual cluster" is
developed.
A mutual cluster is a set of observations whose largest within-group
distance is smaller than the distance to the closest point outside the
set. The hybrid method is a top-down approach that preserves such
mutual clusters.
Mutual clusters are useful in their own right, and have the interesting
property that they cannot be "broken" by many bottom-up methods.
Tools
that graphically integrate mutual clusters into bottom-up methods are
also included in the library.
Lastly, an implementation of Michael Eisen's bottom-up clustering
algorithm is also in the library.
A paper, soon to appear in Biostatistics, describes this work:
Chipman, H., and Tibshirani, R. "Hybrid Hierarchical Clustering with
Applications to Microarray Data", available at
http://ace.acadiau.ca/math/chipmanh/papers/chipman-tibshirani-2003-modif
ied.pdf
There's an online tutorial at
http://ace.acadiau.ca/math/chipmanh/hybridHclust/index.html
Hugh Chipman and Robert Tibshirani
-- Hugh Chipman
-- Associate Professor and Canada Research Chair in Mathematical
Modelling
-- Department of Mathematics and Statistics, Acadia University
-- Wolfville, Nova Scotia, Canada B4P 2R6
-- (902) 585-1525, Fax: (902) 585-1074
-- http://ace.acadiau.ca/math/chipmanh/homepage.htm
_______________________________________________
R-packages mailing list
R-packages at stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-packages