When selecting multiple responses, JMP currently creates an individual Tree for each response. These tree can later be combined for analysis.
Instead, we would like to be able to have a single tree that best fits all the responses simultaneously.
This capability is available as python function call RandomForestRegressor in the the popular Sklearn package.
There are many papers available on the subject, for example:
"Splitting variable selection for multivariate regression trees", Hsiao and Shih, 2006, ScienceDirect.
We would like this capability, once implemented for decision trees, extended to the Bootstrap Forest and Boosted treed methods.
