Describing Categorical Variables
Prediction Analysis (PA) of cross-classifications of categorical variables proceeds from a different set of assumptions. Prediction analysis is performed on sets of point predictions in cross-classifications of predictors and criteria. PA allows one to analyze these hypotheses from two perspectives. The first is to evaluate whether this set of three hypotheses, overall, allows one adequately to describe the data. The second perspective is to evaluate each hypothesis separately. Hypotheses can be properly tested only if they are properly formulated. PA allows certain patterns to be left out from specification of hypotheses. The identification of hit cells and error cells for a PA starts from the cross-tabulation generated to display the joint frequency distribution of all variables involved in a set of prediction hypotheses. PA discriminates between hit cells and error cells. Equifinality models trace back a single outcome to various antecedents. Equicausality models trace back two or more outcomes to one and the same antecedent. .