An improved two-stage variance balance approach for constructing partial profile designs for discrete choice experiments

摘要: In many discrete choice experiments set up for product innovation, the number of attributes is large, which results in a substantial cognitive burden respondents. To reduce such cases, Green suggested early '70s use partial profiles that vary only levels subset attributes. this paper, we present two new methods constructing Bayesian D-optimal profile designs estimating main-effects models. They involve alternative generalizations Green's approach makes balanced incomplete block and take into account fact may have differing numbers levels. We refer to our as variance balance I II because they an attribute with larger more often than fewer stabilize variances individual part-worth estimates. The differ way are weighted. Both provide statistically efficient another generalization does not weight This method called balance. show from actual experiment software development demonstrating usefulness methods. Copyright © 2014 John Wiley & Sons, Ltd.