APA Dictionary of Psychology

n. a procedure used to assess the utility or stability of a statistical model. A data set is randomly divided into two subsets, the first of which (the derivation sample) is used to develop the model and the second of which (the cross-validation sample) is used to test it. In regression analysis, for example, the first subset would be analyzed in order to develop a regression equation, which would then be applied to the remaining subset to see how well it predicts the scores that were actually observed.