Hierarchical Bayes Conjoint Choice Analysis via the Approximate Dependent Poisson Race Model

• Activity Number: 32
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Hang Joon Kim*+ and Steven MacEachern
• Companies: The Ohio State University and The Ohio State University
• Keywords: Poisson race model ; hierarchical Bayes ; choice model ; conjoint analysis ; MCMC

Abstract

Conjoint choice analysis is widely used in many applications, e.g. marketing, economics and transportation. However, the dependence structure between attributes of choice alternatives and different time points makes the modeling and the estimation challenging. A study conducted by Ruan et al. (2007) found that the class of Poisson race models captures the dependence structure of consumer choice and has better performance than the independent Poisson race models and the multinomial logit model. Despite the advantage, discreteness of latent utility variables makes it difficult to apply the model to multiple choices data. We propose the approximate dependent Poisson race model that incorporates the cross-sectional and temporal dependence with multiple alternatives. In addition, we discuss the implementation of an efficient MCMC algorithm that incorporates the complex dependence structure.