A Dichotomization Method for Boolean Analysis of Quantifiable Co-Occurrence Data

Boolean analysis is a partial order generalization of scalogram analysis. It enables one to build a family of implication schemes that bears a 1-1-relationship with a data set. Unlike scalogram analysis, this method does not impose any constraints on the data structure. As a result, implication schemes may become extremely complex and may contain errors. Therefore, methods are needed for the selection of subsets of data to be modeled (dichotomization methods). In this chapter, a dichotomization method is introduced which uses the relative frequency of data patterns with the same number of correct responses as a given criterion. It is argued that the present approach yields implication schemes which comply better with the underlying data structure, especially when the subjects show similar abilities in solving the items.