Re: Cluster : List of Cases

[John Darrington]

I suspect the answer is, that these are options which haven't yet been 
implemented
in the QUICK CLUSTER command.  However, you can probably achieve most of these
things with a bit of manual intervention.  For example, you can get the distance
between a case and the center of a cluster using the COMPUTE command.

I have sent a message to the original author of the QUICK CLUSTER command to see
if he has any other suggestions.

Regarding your question about testing that bot sets of cases come from the same 
population, it seems to me that this is a goodness-of-fit test, and hence 
the Chi-Squared Test would be appropriate.


J'


On Sat, Apr 21, 2012 at 07:57:43PM -0700, stn021 wrote:
     
     Hi, 
     the K-Means-Clusteranalysis computes a number of clusters and says how many
     cases belong to each cluster. 
     
     How can I
     1) find out which cases belong to which cluster?
     2) apply the clusters to other cases that were not included in the 
analysis.
     What I mean is: suppose I have 1000 observations from the same population
     and use 500 of those to calculate the cluster-means, then the other 500
     should fit into the same clusters, otherwise it would not be the same
     population. So how can I
     2.1) match these cases to the clusters and 
     2.2) test the underlying claim that both sets of observations really belong
     to the same population?
     

-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://keys.gnupg.net or any PGP keyserver for public key.

signature.asc
Description: Digital signature