which distance does JMP use in clustering?

Question

Hello everybody.

I'm trying to understand which distance does JMP use in hierarchical clustering: the documentation in https://www.jmp.com/support/help/en/15.1/index.shtml#page/jmp/distance-method-formulas.shtml#ww177809 seems to say "squared euclidean distance" but if I save the distance matrix (after clustering) what I get is the euclidean distance (not squared).

And what about the k-means method? It should use the euclidean distance, but I couldn't find the formulas.

Thank you in advance.

Chris_Kirchberg · Answer

Hi @francesco_della

No problem. The squared distance is used in the subsequent between cluster distance calculations in some methods, but the initial distance matrix between the observations is not squared.

Best,

francesco_della · Answer

thank you so much for your prompt answer, Chris
I was just confused because the formula in the documentation definitely shows the SQUARED euclidean distance
thank you again, regards

Chris_Kirchberg · Answer

Hi @francesco_della ,

JMP uses Euclidean Distance for the initial distance matrix calculation between observations and then the method chosen for calculating distances between the clusters. This is true for any of the Clustering methods within JMP. The only exception is if you provide the data as a distance matrix and choose the data is distance matrix option in the dialog for Hierarchical Clustering.

For k- Means, the JMP help refers to the SAS FASTCLUS Procedure documentation found here.

Hope that helps.