Cross-Validation Assessment of Alternatives to Individual-Level Conjoint Analysis: A Case Study

Hagerty uses Q-type factor analysis and Kamakura a hierarchical cluster analysis that optimizes predictive validity and both approaches are compared with conventional conjoint and self-explicated utility models using real datasets.

Abstract

Recently, both Hagerty and Kamakura have proposed insightful suggestions for improving the predictive accuracy of conjoint analysis via various types of averaging of individual responses. Hagerty uses Q-type factor analysis (i.e., optimal weighting) and Kamakura a hierarchical cluster analysis that optimizes predictive validity. Both approaches are compared with conventional conjoint and self-explicated utility models using real datasets. Neither the Hagerty nor the Kamakura suggestions lead to higher predictive validities than are obtained by conventional conjoint analysis applied to individual response data.