JMP Pro 13 Modeling Workflow Enhancements with GenReg, Screen Predictors, and the new Formula Depot
Karen Copeland, Boulder Statistics
Building models is often an exploratory, iterative process that can result in many saved models as one works to develop the most useful model for a particular application.The new Formula Depot in JMP13 Pro provides a workflow to save and compare models without cluttering your data table with numerous save model columns.
Prior to building models one may need to screen a large number of factors to obtain a list suitable for modeling (e.g., moving from 1000 predictors to say 25 predictors). The JMP model utility screen predictor is one tool that can help reduce the number of potential factors to consider. Once a smaller set of factors is identified then the generalized regression personality in the fit model platform is one tool for model building. Generalized regression options include various algorithms and validation options to facilitate factor selection and/or minimize overfitting. In addition, the generalized regression capabilities have been expanded in JMP Pro 13 both in terms of the types of models available as well as model visualizations, such as ROC curves and profilers now available within the GenReg platform.