A model of ranked conjoint-data and implications for evaluation

This article examines basic features of ranked Conjoint-data, analyzes the adequacy of evaluation methods and proposes improvements for better utilizing the information provided by ranked data. It is shown that commonly used goodness-of-fit measures provide inadequate proxy measures for assessing rank consistency and internal validity of estimates. In addition, commonly used evaluation methods, such as OLS and LINMAP, are shown to be based on arbitrary propositions which do not fulfill the requisite traits postulated by the model of ranked Conjoint-data. Resulting shortcomings on estimation outcomes are evaluated with means of simulation analyses. New insights into the achievable estimation accuracy are gained and possibilities for improvement are shown.