Quoting Micheal Wagner <
[hidden email]>:
> How would I calculate my Sums of squares from a 222x18 matrix with
> 14
> unique values from cluster analysis.
>
> The end objective being to evaluate between and within group sums of
> squares.
>
Run your cluster analysis in the usual way, but use the "SAVE" button to
save "Cluster membership". This will be a new variable with a name like
CLU3_1 (the first analysis and its three-cluster solution).
Then use General Linear Model, select the Multivariate option, select
the variables which went into the cluster analysis, and specify the
cluster membership variable as the fixed factor. The "tests of
between-subjects-effects" output doesn't give you the individual sums of
squares, but it does with them what you probably want, it calculates
mean squares, an F ratio, and a significance value.
The significance value is a useful guide to the quality of the
clustering, but it is not a strictly correct probability value, as you
haven't tested a pre-specified hypothesis, but rather one of the "best"
hypotheses that the cluster analysis could find.
David Hitchin
=====================
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
Locked
