Tree-Based Models

Tree-based models provide an alternative to linear and additive models for regression and classification data. Such models are fit by binary recursive partitioning whereby a dataset is successively split into two subsets until it is infeasible to continue. We discuss an implementation in the new S language [1] which consists of a number of functions for growing, displaying, and interacting with tree-based models. This approach to tree-based models is consistent with the data-analytic approach to other models, and consists primarily of fits, residual analyses, and intense graphical inspection.