Are you interested in the tree itself, or do you
intend to slice it?
Does it happen with each of the proximity-algorithm combinations
you are trying?
Somethings vary according to proximity measure and clustering
algorithm chosen, but a fanciful solution would have zero SD's
within a cluster, i.e., all cases in a cluster would have scores
exactly equal to the centroid.
An off the cuff guess, you have ties.
some checks:
sort the cases according to cluster membership say for
however many clusters you have at 42 steps.
Redo the clustering put the proximity matrix in the listing.
and inspect the distance/similarity coefficients are they very close
to the same?
use MEANS with the cluster membership at that point as the group
variable.
Are the SDs small?
Eyeball the scores themselves for the first few cluster, how similar
do they look to you?
What is the nature of your variables? What is the legitimate score
range of the variables?
What constitutes a case?
Art Kendall
Social Research Consultants
On 4/20/2011 3:14 AM, ilksun ulbegi wrote:
Hi,
I am using cluster analysis for my research, but when I look at
Agglomeration Schedule my coeeficients are all zero for 42
stages, then it begins to rise 0,20; 0,40 etc. Does anybody know
why this happens? I looked many books but no answer yet:(
Ilksun D. ULBEGI
=====================
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
Social Research Consultants
Locked
