Can anyone tell me , how can I refine the already made clusters
(generated by human intelligence) in SAS ? ..... that is, I already
have the clusters but I need to refine those by some variables.

On Dec 30, 3:44=A0am, Pratyush <pratyush...@gmail.com> wrote:
> Can anyone tell me , how can I refine the already made clusters
> (generated by human intelligence) in SAS ? ..... that is, I already
> have the clusters but I need to refine those by some variables.

I think that you'll have to provide an example dataset, preferably in
the form of a datastep, and the output file you want (also, preferably
in the form of a datastep).

Art

On Dec 30, 1:44=A0am, Pratyush <pratyush...@gmail.com> wrote:
> Can anyone tell me , how can I refine the already made clusters
> (generated by human intelligence) in SAS ? ..... that is, I already
> have the clusters but I need to refine those by some variables.

Am I right that you have some form of cluster structure for p
variables and
you have now q new variables added to p+q  variables?
Am I right that you want to do a stepwise algorithm improving the
cluster
structure w.r.t. all p+q variables?
If you do Kmeans or Kmedian you would have to check the cluster
membership with new distances based on the larger set of variables,
compute new cluster centers and check each member if it must
be moved to another cluster.

Of course that would probably not result in a global but hopefully
a local optimal solution (if you are able to avoid cycles).

But wothout knowing what cluster method you have been using,
there is not much to say about this. For example, if you have a
hierarchical structure, I would not think that it would be suitable
and I guess then you would have to run the entire tree again.