On the Use of Prediction Analysis in Cognitive-Developmental Research: Examples from Childhood and Old Age.

This paper describes four research examples of the use of prediction analysis in research. The aim of prediction analysis is to compare a rule as specified by a theory with a reference rule. The first example of prediction analysis in this paper concerns a study of 120 first, second, and third graders' performance on length and weight problems. Results showed that children would not be able to solve a weight problem at a given level if they did not solve the length task of the same and all lower levels. In the second example, the hypothesis that a certain amount of attentional capacity is a necessary condition for attaining certain forms of reasoning was tested and supported. In the third example, children solving a transitivity task were divided into two subpopulations: (1) an operational subpopulation, in which children infer judgments on the basis of premise relations; and (2) a nonoperational population, in which children infer judgments in some other way. The hypothesis that only the combination of these two populations produces stochastic dependence was supported. Data from the fourth study, whose 80 subjects had a mean age of 81 years, showed that it may not be meaningful to recode a continuous variable into two or more categories in order to apply prediction analysis. (SH)