Latent Class Models for the Analysis of Behavioral Hierarchies

The author develops and describes latent class models useful for the analysis of behavioral hierarchies. The latent class models investigated are generalizations of the Guttman scale model and consider probabilistic relationships of item response to scale type and multiple hierarchical orderings of item responses. In addition, the author develops models for hierarchies that are present at the level of sets of items rather than at the level of individual items. He calls them “characteristics models” and examines their relationship to models for hierarchies of items. The models are illustrated on consumer complaint data gathered from a cross-sectional survey.