Dear allstat - members with experience in cluster analysis,
I want to perform a cluster analysis on psychometric data consisting of 35 variables, abot 1/3 of which are continuous, binary and ordnial.
The only widely accepted method to combine variables of this type is, to my knowledge, the Gower's coefficient.
As probably not all of the variables contribute to reasonable clustering, I have to perform a variable selection.
Is anyone of you familiar with the problem of selecting variables
for clustering based on a Gower's similarity or distance matrix?
Kind regards
Rainer Grauf
email:
[log in to unmask]
stop
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|