=========================================================================
Date:         Mon, 31 Jul 2006 19:22:10 +1000
Reply-To:     paulandpen@optusnet.com.au
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Paul Dickson <paulandpen@optusnet.com.au>
Subject:      Re: Distance from cluster centre query.
Comments: To: Mark Webb <targetlk@iafrica.com>
Content-Type: text/plain

Mark

I do not think this is possible using K-means due to the algorithm used, but I may be wrong.  One way round it might be to work out which variables are contributing to the cluster solution.  Then formulate an algorithm based on Chaid, Cart or Discrim, Logistic and assign each case a score using the algorithm (coding rules).  You could then compute scores from the algorithm and for all cases that are assigned to that cluster with a high (say 80%) level of probability generate means and standard deviations and treat the mean scores as cluster centres and the standard deviations as your index of dispersion (i.e. distance from cluster centre).

Cheers Paul

> Mark Webb <targetlk@iafrica.com> wrote:
>
> In K Means it's possible to save this information as a variable.
> Is this possible in any of the hierarchical methods offered in SPSS ?
> They offer a proximity matrix - which I see as different - as this shows
> distances between individual respondents NOT the classification mean.
> Am I missing something ?
>