A statistical approach to assessing interval scale preferences in discrete choice problems

This paper introduces two statistical interval scale techniques for assessing preferential uncertainty in discrete choice problems. The main benefit of interval scale techniques is that the evaluations can be done by local best–worst scale so that there are concrete reference points defining the utility scale. However, since value measurement is based on the decision–makers' subjective preferences, information is also subject to uncertainty. For example, decision–makers may have difficulties in evaluating certain alternatives or they are not necessarily consistent with their own pairwise evaluations. The proposed methods differ from previous interval scale methods in that also the decision–maker's uncertainty is evaluated. The advantage of the statistical approach is that it permits inconsistency in evaluations and can thus lead to more realistic description of the overall preferences when compared to deterministic approaches. Furthermore, the statistical approach permits the examination of the impacts of uncertainty on overall decision making.