hierarchical bayesian model development with applications in marketing

Open Access
Chen, Zhe
Graduate Program:
Doctor of Philosophy
Document Type:
Date of Defense:
August 08, 2011
Committee Members:
  • Duncan Fong, Dissertation Advisor
  • Duncan Fong, Committee Chair
  • Rana Arnold, Committee Member
  • Murali Haran, Committee Member
  • Ping Xu, Committee Member
  • Wayne Desarbo, Special Member
  • hierarchical bayesian model
  • MNP model
  • MDS
Hierarchical Bayesian models have been widely used in practice to address different kinds of problems in many disciplines. Typically, efficient Markov Chain Monte Carlo simulation methods are developed to generate random samples from complicated posterior distributions resulting from such models. In the Marketing literature, many hierarchical Bayesian models have been devised to investigate various marketing phenomena. In this dissertation, I provide new hierarchical Bayesian models to investigate three marketing problems which have not been well studied in the literature, namely: (1) a heterogeneous Bayesian dynamic model to study the association between customer satisfaction and a firm’s financial performance, (2) a Bayesian random-coefficient multinomial probit model to analyze customer choice panel data and evaluate the value of purchase history data in direct marketing, and (3) a Bayesian vector multidimensional scaling with variable selection procedure to produce a joint space map of consumers and brands for product positioning analysis. In addition to developing Bayesian models for these problems, we apply the methods to analyze real data and draw insights from our Bayesian analysis with marketing applications. Simulation studies with comparison results of some benchmark models will also be presented.