Classification and Regression Trees

These are nonparametric procedures for explaining and/or predicting a response, either categorical (then this is discriminant analysis or classification), or continuous (then this is a nonparametric regression).

The explanatory variables used may be both categorical and continuous, and it is this that makes the method useful in contexts that would be too complex to use analysis of covariance, for instance.

We will detail the classification case more as the risk and criteria used are less obvious than in regression, where least squares is as usual the staple loss.