How to get Cluster Centers for Hierarchical Cluster Analysis SPSS

[us...@domain.invalid]

I am conducting cluster analysis with SPSS and was wondering how (if
possible with SPSS) to obtain and save cluster centers (not membership)
for Hierarchical Cluster Analysis?

These cluster centers would be saved in a separate file,
whereby they would be used as seed points to conduct Kmeans Cluster
Analysis.

Thank you in advance.
Paul

[Richard Ulrich]

It sounds to me like you are talking about taking
one level of hierarchy, sorting on the Group,
and aggregating to get the mean of the variables.

Or else, selecting the individuals that are most
central, assuming there is some measure of that
in the hierarchical analysis.

Does that describe what you need?

--
Rich Ulrich, wpi...@pitt.edu
http://www.pitt.edu/~wpilib/index.html

[us...@domain.invalid]

Possibly your first response.

In running the Hier CA, I ask for 3 cluster solution and save cluster
membership. But how can I save the cluster center (mean score?) of each
variable for each cluster? (similar to Kmeans where you save cluster
centers)

So for 10 variables and 3 clusters there would be 10 means (cluster
centers) = 30 mean scores.

I guess what I could do is split the file by cluster, 3 samples, and
calculate the mean of each variable, then would that be the cluster
center? But the cluster center is based on Euclidean distances
(usually), not raw scores, so then the above wouldn't work...?...

[Paul Spanier]

Possibly your first response.

In running the Hier CA, I ask for 3 cluster solution and save cluster
membership. But how can I save the cluster center (mean score?) of each
variable for each cluster? (similar to Kmeans where you save cluster
centers)

So for 10 variables and 3 clusters there would be 10 means (cluster
centers) = 30 mean scores.

I guess what I could do is split the file by cluster, 3 samples, and
calculate the mean of each variable, then would that be the cluster
center? But the cluster center is based on Euclidean distances
(usually), not raw scores, so then the above wouldn't work...?...

[monia9PL]

If you want to have all these means in one table in order to export it,
use Custom Tables. This will also allow you do make a multiple
comparison of means of each variable ("Test Statistics") to describe
the groups.

Hope helped
Monika

Paul Spanier napisal(a):

> begin:vcard
> fn:Paul Spanier
> n:Spanier;Paul
> org:University of Toronto;Faculty of Physical Education & Health
> adr:55 Harbord Street ;;Graduate Deptartment of Exercise Science ;Toronto ;Ontario;M5S 2W6;Canada
> email;internet:paul.s...@utoronto.ca
> title:PhD Candidate
> tel;home:905-388-4319
> tel;cell:905-741-9103
> url:http://individual.utoronto.ca/EPU/index.htm
> version:2.1
> end:vcard

[Richard Ulrich]

On Fri, 02 Dec 2005 14:36:05 -0500, Paul Spanier
<paul.s...@utoronto.ca> wrote:

> Possibly your first response.
>
> In running the Hier CA, I ask for 3 cluster solution and save cluster
> membership. But how can I save the cluster center (mean score?) of each
> variable for each cluster? (similar to Kmeans where you save cluster
> centers)
>
> So for 10 variables and 3 clusters there would be 10 means (cluster
> centers) = 30 mean scores.
>
> I guess what I could do is split the file by cluster, 3 samples, and
> calculate the mean of each variable, then would that be the cluster
> center?

That's what is accomplished by what I wrote about
sort-then-Aggregate.

> But the cluster center is based on Euclidean distances
> (usually), not raw scores, so then the above wouldn't work...?...

What do you think that a Euclidean distance is?
The distance can be computed using the raw scores or
the standardized versions of the raw scores -- either
way, the center - in terms of the raw scores - is the
mean of those scores. Isn't it?

[snip, previous]