Model fusion of results from disparate survey methodologies is a topic of current interest in both research and practice. Much of this interest has centered on the enrichment of stated-preference results with revealed-preference data, or vice versa, as it is considered that stated preference methods provide more robust trade-off information while revealed preference methods give better information about market equilibria. The motivation for this paper originates in the automotive industry, and is distinct in that it focuses on the reuse of existing data. Practitioners wish to glean as much information as possible from a large body of existing market research data, which may include minimally overlapping datasets and widely varying survey types. In particular, they wish to combine results from different types of stated preference methods. This paper presents two advancements in model fusion. One is a method for reducing data gathered in open-ended methods such as van Westendorp studies to a form amenable to analysis by multinomial logit, thus enabling the comparison of open-ended data to conjoint data on overlapping data sets. The other is a new statistical test for the fusibility of disparate data sets, designed to compare different methods of data comparison. This test is less sensitive than existing tests, which are most useful when comparing data sets that are substantially similar. The new test may thus provide more guidance in the development of new methods for fusing distinct survey types. Two examples are presented: a simple study of cell phone features administered as a test case for this research using both choice-based conjoint and van Westendorp methodologies, and a pair of existing larger-scale studies of automotive features with some attributes common to both studies. These examples serve to illustrate the two proposed methods. The examples indicate both a need for continued testing and several potentially fruitful directions for further investigation.

This content is only available via PDF.
You do not currently have access to this content.