Skip to content

cluster-analysis and NA's

2 messages · chr.schulz@email.de, Christian Hennig

Original
chr.schulz@email.de
# 
Hi,

exist a special cluster-analysis algorithms
which can work with NA's.

a further "problem" is that i want cluster
variables not cases to identify special variable-set's.

Is it a common way turn the data.frame and use
kmeans,because this works with NA's, or have anybody another
method for finding "variable-sets"  , with exception of factor analysis.

thanks for advance
Christian

 

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Christian Hennig
# 
Hi,
On Wed, 7 Aug 2002 chr.schulz at email.de wrote:

            

      
      
      
    

        
The choice of a clustering method depends very much on the aims of your
analysis and especially on your concept of "similarity" between the
entities to cluster. One of the most often used methods for clustering
variables is to compute a dissimilarity matrix between the variables, where
dissimilarity is defined as 1-correlation or 1-|correlation| (depending on
if you want variables pointing in almost exactly the opposite direction 
to be interpreted as similar or as extremely dissimilar).
Then you can apply one of the algorithms of hclust or agnes of library
cluster (the latter is not a bad choice as a default). 
NAs are not a big problem because correlations can be properly computed in
presence of NAs.

Christian