An empirical comparison of maximum difference sensitivity modeling and discrete choice analysis.

摘要

The focus of this dissertation is to investigate maximum difference judgment analysis, to make several comparisons between maximum difference judgment analysis and discrete choice analysis, and to introduce sensitivity analysis. The maximum difference judgment procedure requires that a set of profiles be generated. The task of the respondent is to choose a pair of attribute levels within each profile which are the most distinct; that is, select the attribute at a prespecified level that makes the profile most attractive and the attribute at a prespecified level that makes the profile least attractive on some specific dimension of interest. In discrete choice analysis, several profiles are generated for each choice set. The task is to select the one profile in each set that is the most preferred. One can see that these two methods model different tasks and an empirical comparison suggests that a strong relationship does not exist between these two methodologies. However, using a combination of maximum difference and discrete choice analyses can yield significant insight into the underlying behavioral action being studied and provide a powerful tool to the decision maker. Finally, sensitivity modeling, which refers to a family of techniques which will enable one to model respondents' preferences in such a way that interdimensional comparisons can be made between attributes, will be introduced. Several sensitivity models will be suggested as alternative methods to maximum difference judgment analysis.