[R] clustering dissimilarities

[Huntsinger, Reid]

The model-based methods I'm familiar with (mixture models) require
the "ambient space" from which the sample was drawn, not just the
metric restricted to the sample points. You could try an embedding
approach like multidimensional scaling (e.g. cmdscale in package mva)
with fairly high dimension and then use a model-based approach on the
result. The choice of embedding will likely have some influence on
the final result.

Reid Huntsinger

-----Original Message-----
From: Don Eduardo Miranda [mailto:miranda at di.fct.unl.pt]
Sent: Friday, December 13, 2002 4:28 PM
To: r-help at stat.math.ethz.ch
Subject: [R] clustering dissimilarities


Hello.  I know my dissimilarity matrix but not my original data.  Is there
any way i could use the clustering function Mclust or EMclust with this
dissimilarity matrix?   or at least some equivalent of these functions?  As
this is model based clustering i dont know if it is actually possible to do
it without the original data


thanks in advance for your help

	[[alternate HTML version deleted]]

______________________________________________
R-help at stat.math.ethz.ch mailing list
http://www.stat.math.ethz.ch/mailman/listinfo/r-help

------------------------------------------------------------------------------