Uniqueness Issues in the Decompositional Modeling of Multiattribute Overall Evaluations: An Information Integration Perspective

The author reviews how methods developed within the information integration paradigm can be used to study consumers’ overall evaluations of choice alternatives. Methods are presented for determining the adequacy of several common model forms used to represent overall evaluations: adding, multiplying, and multilinear. Often, more than one integration model can be reconciled with the data by altering one's assumptions about the subjective values of the independent variables and about the relationship between private, unobservable overall evaluations and the overt numerical ratings that index them. Also, different integration models lead to parameter estimates (e.g., part worths) of varying levels of uniqueness and interattribute comparability. Emphasis is given to pinpointing the sorts of evidence and experimental designs that enable one to distinguish empirically among alternative model forms and psychological interpretations of the data—and, conversely, to what interpretations cannot be distinguished empirically—given only overall evaluations of a set of choice alternatives that vary along two or more attribute dimensions. Finally, the methods described are compared with model diagnosis procedures more commonly used in marketing and consumer research, including compositional correlational techniques and decompositional methods of conjoint measurement.