The JMP Partition platform (Analyze > Predictive Modeling > Partition) creates a decision tree by splitting (partitioning) data according to a relationship between the predictor and response values. JMP searches all possible splits of predictors to best predict the response. The splits continue until the desired fit is reached. JMP generates different reports according to the type or response (Continuous or Categorical) and a Candidate list node report for both.
Alternately, or in combination with the automatic splits, you can choose to split on any term. For Categorical responses, you may want to choose a term from the Candidates list by examining the G^2, the likelihood ratio chi-square for the best split. Likewise, for Continuous responses you can examine the sum of squares to help determine the best split.
JMP Partition Decision Tree
Partition Candidates report for Categorical responses
Partition Candidates report for Continuous responses
Technical Enablement Engineer Peter Hersh @Peter_Hersh gives a brief overview of how to force a split.
