Imprecise probability and expert forecasting

Abstract

Evidence is often insufficient to support the assessment of precise probabilities. Shifting to vaguer measures of uncertainty, such as upper and lower probabilities, does not deprive one of the key analytical tools of classical probability. Two approaches to the calculation of upper and lower expected values are described and contrasted in the case of forecasting production costs of an electric utility. Conditionalization of imprecise probabilities is also discussed.<>