Date:         Tue, 11 Apr 2006 12:05:53 -0400
Reply-To:     Leyla Yerlikaya <leylay@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Leyla Yerlikaya <leylay@HOTMAIL.COM>
Subject:      Re: Using Varclus for PreDetermined Clusters

On Tue, 11 Apr 2006 10:54:21 -0400, Edzard van Santen
<evsanten@ACESAG.AUBURN.EDU> wrote:

>Assuming that you have multivariate data why not used Proc Candisc with
the distance option and
>the clusters created as a class variable to calculate Mahalanobis
distances?
>
>
>On Tue, 11 Apr 2006 10:22:28 -0400, Leyla Yerlikaya <leylay@HOTMAIL.COM>
wrote:
>
>>Hello:
>>
>>I have created 4 clusters using Proc Cluster. I want to somehow force the
>>same 4 cluster solutions using Proc Varclus in order to see the R-sqr with
>>Own Cluster and R-sqr with Next Closest. Is there anyone who knows how to
>>force feed a cluster solution to Proc Cluster?
>>
>>Thanks,
>>Leyla Yerlikaya

Candisc gives the squared distances between clusters at an aggreagte level.
I need to see each observation's "closeness" to its own cluster and to the
next cluster.