Large sample properties of cross-validation assessment statistics (Q1094026)

Asymptotic properties are developed for two classes of cross-validation assessment statistics. The two correspond to a large number of small partition sets and a small number of large partition sets. Conditions implying weak convergence, strong convergence, and asymptotic normality are presented for each class. A discriminant analysis misclassification estimator and a regression analysis estimator of absolute error are used to illustrate the general theory.