Abstract:

This paper compares conjoint and logit modeling approaches/ when input data is a complete rank ordering of the concepts in a full factorial design, and when input data is a rating scale (i.e. an ordered categorisation) of the concepts in a full factorial design. The comparisons are made not only in terms of how well each model fits the judgmental data on the experimental concepts, but also in terms of how well each of them predicts the judge's evaluations of a second set of hypothetical concepts. The comparisons indicate that logit modeling performs almost as well as traditional conjoint analysis in terms of both internal and predictive validity.