Setting Seeds in K-means

[Peter Brooks-6]

Hi

I'm looking to cluster 50 subjects I am testing, analysing 40
dependant variables. I am clustering initially using hierarchial
analysis (producing 5 clusters) but want to validate the 5 clusters
with K-means. Accordingly I want to use the means of each variable, in
each cluster from the initial clustering, as seeds in K-means but no
matter how I try I can't get SPSS to read the file I produce.
Can someone help me?

Peter Brooks
MPhil/PhD Student
Centre for Sport and Exercise Sciences
University of Greenwich

Please consider the environment before printing
--------------------------------------------------------------------
University of Greenwich, a charity and company limited by guarantee,
registered in England (reg. no. 986729).  Registered office:
Old Royal Naval College, Park Row, Greenwich, London SE10 9LS.
--------------------------------------------------------------------

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Art Kendall]

Starting with a hierarchical analysis is a good place to start.
However, rather than using K-means, I would suggest using TWOSTEP.  It
gives a lot more information for what it sounds like you are trying to
do. It includes a hierarchical analysis and AIC or BIC for different
numbers of clusters.  These stats are useful in ball-parking the number
of clusters to retain. It also gives info about the influence of
variables, etc.

Are the 40 variables from established scales? Are they some other form
of repeated measures?

Also it would be helpful if you posted the syntax you are using and the
error messages you are getting.


Art Kendall
Social Research Consultants

On 1/27/2010 9:31 AM, Peter Brooks wrote:

> Hi
>
> I'm looking to cluster 50 subjects I am testing, analysing 40
> dependant variables. I am clustering initially using hierarchial
> analysis (producing 5 clusters) but want to validate the 5 clusters
> with K-means. Accordingly I want to use the means of each variable, in
> each cluster from the initial clustering, as seeds in K-means but no
> matter how I try I can't get SPSS to read the file I produce.
> Can someone help me?
>
> Peter Brooks
> MPhil/PhD Student
> Centre for Sport and Exercise Sciences
> University of Greenwich
>
> Please consider the environment before printing
> --------------------------------------------------------------------
> University of Greenwich, a charity and company limited by guarantee,
> registered in England (reg. no. 986729).  Registered office:
> Old Royal Naval College, Park Row, Greenwich, London SE10 9LS.
> --------------------------------------------------------------------
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Greg James-2]

Peter,

 

Is there a possibility that across all the DV's, there are enough missing variables to ensure that no one record is complete? If so, I'm betting SPSS is throwing out every observation without complaint. I ran across the same behavior in Clementine a while back.

 

-Greg James

-----Original Message-----
From: Peter Brooks [mailto:[hidden email]]
Sent: Wednesday, January 27, 2010 09:31 AM
To: [hidden email]
Subject: Setting Seeds in K-means

Hi I'm looking to cluster 50 subjects I am testing, analysing 40 dependant variables. I am clustering initially using hierarchial analysis (producing 5 clusters) but want to validate the 5 clusters with K-means. Accordingly I want to use the means of each variable, in each cluster from the initial clustering, as seeds in K-means but no matter how I try I can't get SPSS to read the file I produce. Can someone help me? Peter Brooks MPhil/PhD Student Centre for Sport and Exercise Sciences University of Greenwich Please consider the environment before printing -------------------------------------------------------------------- University of Greenwich, a charity and company limited by guarantee, registered in England (reg. no. 986729). Registered office: Old Royal Naval College, Park Row, Greenwich, London SE10 9LS. -------------------------------------------------------------------- ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD