Using Basic Experimental Design Principles to Design Better Conjoint Studies

• Activity Number: 280
• Type: Contributed
• Date/Time: Wednesday, August 14, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics and Marketing
• Author(s): Steven LaLonde*+
• Affiliation(s): Rochester Institute of Technology
• Keywords: conjoint analysis ; design efficiency ; mixed level designs ; experimental design ; orthogonal designs

Abstract

Using statistically-designed experiments has become increasingly popular in market research. However, basic design principles are often underutilized in market research. The design of an experiment undertaken in research and development often involves meticulous consideration of the number of factors, the number of levels of each factor, the possible interaction between factors, and the minimum number of runs sufficient to obtain valid results. In my experience, experimental design in market research focuses more on determining the factors and levels of factors of interest, followed by PROC OPTEX in SAS to determine which of all possible designs of sixteen runs is "best." Less consideration is given as to how many runs would be required to produce an orthogonal design, or the possibility of interactions between factors. Trade-offs between different designs with one more, or one less, factor and between designs with one more, or one less, level-within-factor are often not considered. This paper will demonstrate, through examples, the benefit of taking more time, and applying more expertise, to the design of experiments in market research.