INFORMATION-THEORY AS A UNIFYING STATISTICAL APPROACH FOR USE IN MARKETING-RESEARCH

Information theory is shown to provide a unified approach to a wide range of problems in marketing research. For instance, this approach can be used to obtain characterizations parallel to those of the Hendry system and other entropic approaches with great economy of assumptions and with the added flexibility that constraints can be easily identified for explicit consideration and implemented as needed. Goodness-of-fit tests and decision modelling structures are supplied from these same stochastic models with a range of applications that include market segmentation and brand shifting choices. A basic approach to these and other procedures is therefore obtainable from information theoretic methods. These methods can be used to address stochastic model selection problems and other probabilistic models of marketing choice - for example, Minimum Discrimination Information (MDI) estimation, Logit, Multiplicative Competitive Interaction (MCI), and other important choice models are also shown in this paper to arise naturally from information theoretic formulations with duality relations developed by Charnes and Cooper providing additional simplifications and interpretations.