A New Class Of Bayesian Segmentation Methods For Deriving Heterogeneous Key Drivers Of Service Quality Evaluations

Open Access
Kim, Sunghoon
Graduate Program:
Business Administration
Doctor of Philosophy
Document Type:
Date of Defense:
Committee Members:
  • Wayne Desarbo, Dissertation Advisor
  • Duncan Fong, Committee Chair
  • Dennis Kon Jin Lin, Committee Member
  • Murali Haran, Special Member
  • Johann Baumgartner, Committee Member
  • Bayesian Regression
  • Market Segmentation
  • Variable Selection
  • Managerial Constraints
  • Finite Mixture Models
  • Heterogeneity
This dissertation proposes a series of unconstrained and constrained Bayesian finite mixture regression models tailored to examine heterogeneous response patterns in service quality evaluations by simultaneously identifying the underlying market segments of consumers (heterogeneity) and the differential significant drivers in their evaluation judgments (variable selection), while enforcing various managerial and theoretical implementation restrictions (constraints) into the model. After providing research motivation in a service quality evaluation context, I will review relevant literature of service quality evaluation and segmentation models in Chapter 2. Following the literature review, I will describe the technical details of the unconstrained Bayesian finite mixture regression model with variable selection in Chapter 3 (Kim, Fong, and DeSarbo 2012) and three additional specialty models in Chapter 4 in a SERVQUAL/SERVPERF context. In Chapter 5, a Monte Carlo analysis with synthetic data will demonstrate that the various Bayes regression models can be gainfully employed in identifying and representing heterogeneous response patterns, and that the proposed models are more robust against multicollinearity than existing methods. In Chapter 6, I will illustrate the usefulness of these proposed models using a SERVPERF survey of the National Insurance Company’s customers. Finally, I discuss the limitations and future direction of this research.