How do I bin/normalize data to minimize intra/inter subgroup variance?

paul_h_anderson

 

I have output test data with high variance to mean ratio. This data is correlated to several categorical variables. I would like to normalize the output data into small number of subgroups to minimize the variance to mean ratios. I know I can pick the subgroups and JMP will calculate intra-subgroup and inter-subgroup variance. If I pick good subgroup factors to normalize, then most of the variance is inter-group and much less intra-group. Is there a process that will explore a large number of factors and show intra to inter variance, so I can select an optimum subgroup? Thanks.

 

Kevin_Anderson

Hi, Cousin Paul!

I hate to start a message apologizing for it's possible ignorance, but without more context, it's kind of a shot in the dark...

Have you explored the Partition platform?  It's binary recursive partitioning analyses will certainly "bin" your large-number-of-factors data for you in a manner that minimize the intragroup heterogeneity and maximizes intergroup heterogeneity between the two (and only two!) groups at each split.

This may not be exactly what you want, but it is certainly easily done!