I seem to recall using a trick to get profiles
graphs of cluster centroids, but I cannot recall which graph I
started with on the menu and then tinkered
Profile graphs are a very old-fashioned way of presenting what
clusters are. They are also useful visualizations for manova or
repeated measure on a common measurement scale. The scale might be
Z-scores, percentages, Likert scales on 1 to 5, etc.
The current problem:
I have 14 variables all on the same scale, zero to one.
I have calculated 5 cluster centroids via AGGREGATE.
I know that the trick was to add a file to the aggregated file
with 2 dummy cases that represent make believe clusters 6 and 7.
One with all zeros and one with all one.
I want to end up with 14 vertical axes with reference lines all
the way across at .25 .50 and .75.
Each cluster would then have a different line style/color. The
lines would connect the means for a cluster across the 14 vertical
axes with markers only at the leftmost and rightmost vertical
axes, i.e. on the outside.
Art Kendall
Social Research Consultants
=====================
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
