An Overview of Some Bayesian Models of Heterogeneity in Marketing

Abstract:
Marketers have long been interested in understanding the differences (in attitudes, behavior etc) between individuals and groups towards products of interest. Hence in Marketing, identifying and evaluating heterogeneity among units (individuals, groups, organizations etc) has been central to research. Heterogeneity models in Marketing can be estimated using either classical or frequentist techniques or Bayesian methods. The latter presents some advantages (especially in small-sample conditions and given latent variable presence). Within the Bayesian paradigm, the widespread implementation of Markov-Chain Monte-Carlo (MCMC) methods in the past two decades has made Hierarchical Bayesian modeling both accessible and fast. We overview some common Bayesian methods that use the Gibbs sampler to assess and estimate heterogeneity in Marketing models. We begin with the simples case of no-heterogeneity in a non-hierarchical model and trace the evolution of more complex models as successive assumptions are relaxed. Associated WinBUGS code to implement the models discussed is also presented