A method for elicitation and combination of imprecise probabilities: A mathematical programming approach

In this paper, a method of elicitation and expert aggregation, supported by existing models in the literature, is developed. To avoid cognitive problems resulting from numerical judgment, comparative judgment of probability is used in the elicitation process, where the best way to represent the data is admitted to be by imprecise probability models. In this sense, a linear programming model is used to convert such judgment into probability intervals. For the aggregation of experts, a quadratic programming model is used, where the use of different metrics to represent the weights will be discussed.