Dispersion binary measure (PROXIMITIES) - error in output/save/cluster

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Dispersion binary measure (PROXIMITIES) - error in output/save/cluster

Kirill Orlov
This post was updated on .
Dispersion similarity is is one of proximity measures for binary data.
Available under Distances menu and PROXIMITIES command. As well as in hierarchical cluster (CLUSTER).
According to the formula and description in the documentation
(Algorithms, Syntax reference) it is a similarity. (Akin covariance,
with some positive values on the diagonal of the matrix).

Well, the error in SPSS (at least version 22) is that in output, the
measure is called "dissimilarity". Moreover, in the matrix saved as
dataset, the ROWTYPE_ value label is PROX DISSIMILARITY, which, since it
is incorrect affects hierarchical clustering (CLUSTER command) that is
sensitive to that label. Results of CLUSTER based on Dispersion
coefficient are incorrect because the program takes it for
dissimilarity. The label must, obviously, be PROX SIMILARITY. It is
similarity measure, like, say Phi correlation coefficient.

So, users, please be attentive.
IBM SPSS team members who visit this site, please consider my note.

----------
ADDENDUM later. I was just informed that the message has been forwarded to IBM SPSS team for consideration.

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (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
Loading...