Robustness of Conjoint Analysis: Some Monté Carlo Results

Abstract

In many industrial applications of conjoint analysis the use of nonmetric algorithms to analyze respondent ranks of products described by more than eight or 10 attributes is time consuming and very expensive for large samples of consumers. The authors compare the results using nonmetric analysis, full factorial designs, and rank data with quicker and less expensive methods of metric analysis, orthogonal arrays and stimulus ratings. In addition, two types of models and levels of error are investigated. The results indicate that metric analysis using ratings data and orthogonal arrays is very robust.