Conjoint analysis by merging attributes

Abstract

Purpose A large number of attributes with mixed levels are often considered in the conjoint analysis. The respondents may have difficulty with scoring their preferences accurately because of many attribute items involved in each survey question. We research on the technique for reducing the number of attribute items. Methods In order to reduce the number of attribute items in a survey question, we make a new attribute by merging two original attributes. A ‘No question’ option is also included as a new level in a merged attribute. Results We propose BIB 64 design in the case where we have four attributes with 2 levels and 3 levels, respectively and then analyze all the respondents survey data generated by the repeated simulation study in order to compare various model selection methods. Conclusion How to reduce the number of attribute items is proposed and how to design and analyze the survey data are illustrated. Key words: Conjoint analysis, Merging attributes, Variable selection methods and model seletion criteria, BIB fractional factorial design