Hierarchical clustering with chi2 distance?

Question

Hi,

A French customer would like to do a hiearchical clustering on his dataset with chi2 distance? I didn't find any reference in documentation about this distance for hierarchical clustering. Perhaps it is defined with another name? is anybordy would know the method which can be consider as similar in JMP?

Thank a lot,

Florence

Chris_Kirchberg · Answer

Yes, please add this as an enhancement request.  SAS PROC DIST allows many distance measures (This is available in JMP Genomics by the way for this very reason). Other serious clustering tools allow one to choose from a list as well (even freeware tools).

jiancao · Answer

I’ve had people asking about using chi squared distances in clustering. It doesn't appear that you can save those from MCA to be used as input data in JMP’s hieratical clustering. I found SPSS’ Clustering Analysis has the option to choose from a variety of measures. So this maybe a feature to add to JMP in the future?

Mark_Bailey · Answer

Multiple Correspndence Analysis is based on Chi square distances. These distances are transformed to get the final coordinates, which you can save. That way might help. The MCA plot is used to find associations between observations (rows).