interpretation of centroid table/plot view when using mixed distance function

[csoares]

Hi
When using kmedoids with a mixed distance function, how to interpret the centroids for the nominal variables? Is it some kind of mean distance?
Thanks and  best regards,
Carlos

[Tobias Malbrecht]

Hi Carlos,

well the nominal distance between equal values is zero - between differing values it is one. Hence for the nominal variables, the centroid is that point, that is the most equal to the other points - which has the most matches among the nominal variables.

Best regards,
Tobias

[csoares]

Thanks, Tobias.

But the centroid table is coding the polynomial variables into numbers (which probably represent the indices of the corresponding value).

Is there a way to convert them automatically back to the original (polynomail) values, to make the centroids easier to read?

Carlos

[csoares]

In case somebody is also looking for the answer for this (which is actually quite simple): operator Extract Cluster Prototypes.

Carlos