Fusion, Pro and Con: Model-Based Fusion of MaxDiff and Ratings Data to Put Respondent Choice Parameters on a Common Scale and Consistency with Random Utility Theory

• Activity Number: 381
• Type: Contributed
• Date/Time: Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Peter Lenk*+ and Lynd Bacon
• Companies: The University of Michigan and Lynd Bacon Associates
• Keywords: Bayesian Inference ; Random Utility Theory ; Psychometric Scaling ; Ordinal Data ; Marketing ; Missing Data

Abstract

MaxDiff scaling, sometimes called Best/Worst analysis, is a special type of choice-based conjoint analysis that involves having respondents pick "best" and "worst" items from choice sets produced from an experimental plan. A weakness of MaxDiff as it is typically applied is that scale location differences between subjects are lost. This shortcoming can be assuaged by augmenting the choice data with information about scale origin, like ratings data. There are different ways of doing this, but not all are consistent with Random Utility Theory (RUT), a cornerstone of preference modeling. In our talk we build on our 2006 JSM results to describe specifications consistent with and not consistent with RUT, compare results from applying them to those obtained from conventional procedures, and indicate the conditions under which the different specifications are to be preferred.