Multitrait–multimethod matrices in consumer research: Critique and new developments

Abstract

In this article we address problems with recently proposed methods for examining construct validity and we introduce alternatives. With respect to the firstorder confirmatory factor analysis model, shortcomings in the application of the method to multitrait‐multimethod data are considered. The correlated uniqueness model is then presented as an alternative when ill‐defined solutions arise in first‐order confirmatory factor analyses or when method and trait variance are confounded. Procedures are then developed for investigating important hypotheses in the application of the direct product model. These procedures were not considered in previous treatments. Finally, panel models are shown to be adaptable to certain investigations of construct validity.