Configural Frequency Analysis Under Multinormality

Configural Frequency Analysis (CFA) searches for local deviations from probability models, the CFA base models. Thus far, base models were specified either in terms of variable relations or in terms of functions that represent data generation processes. In this chapter, these two approaches are fused. Specifically, CFA base models are proposed in which variable relations are incorporated in tandem with cell probabilities that are based on the hypothesis of multinormality. In data examples, it is shown that different sets of cells deviate when either only variable relations or only multinormality characteristics are modeled. When the corresponding base models are fused, these deviations disappear. Extensions of the new approach are discussed.