Uses and computation of imprecise probabilities from statistical data and expert arguments

Imprecise probabilities and the theory of coherent previsions offer a rigorous and powerful framework for modelling subjective uncertainty and solving problems of statistical inference, decision making or risk analysis. The paper introduces formulas for computing imprecise probabilities when statistical data and expert arguments are available to a subject. We then show how to use these imprecise probabilities (dialectical probabilities) for comparing the likelihood of events, conditioning on events, comparing decisions, computing optimal decisions or assessing financial risk. We apply the method to stock trading and show in this experiment the added value both of imprecision and of expert arguments derived from Technical Analysis.