Efficient Designs for a Noncompensatory Choice Model

• Activity Number: 21
• Type: Topic Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Qing Liu*+
• Companies: University of Wisconsin-Madison
• Keywords: experiment design ; conjoint choice design ; semi-Bayesian design ; non-compensatory choice model ; D-optimality ; compensatory choice model

Abstract

Existing research on choice designs has focused exclusively on compensatory models and little guidance is currently available on how to design efficient choice experiments when consumer decision process may be non-compensatory. In this paper we propose a method for constructing efficient choice experiments where consumers may use non-compensatory screening rules. We derive local and semi-Bayesian design criteria for the fairly general non-compensatory conjunctive screening model and identify the characteristics of designs efficient for this model. We find that optimal non-compensatory designs are quite different from compensatory designs. Our results indicate that optimal compensatory designs, when the underlying choice process is non-compensatory, lead to significant loss in estimation efficiency. Readily available and widely used orthogonal designs also perform poorly.