engleski

Discrete Variables with Symmetric and Asymmetric Uncertainties

The paper addresses the issue of multiple information sources and their incorporation in statistical decision models. It is argued that decision makers, when faced with multiple sources of information, update their prior beliefs to form a posterior belief structure later used in the actual decision process. A convenient way of formally incorporating multiple information sources, seen as different (subjective or objective) probability distributions is through the use of the Bayes theorem for events. Hence, a set of formulas for performing Bayesian sensitivity analyses is developed for cases of symmetric and asymmetric uncertainties which allow for combining prior and likelihood distributions in sensitivity analysis on probabilities, while brief discussion is also devoted to sensitivity analyses for importance weights for which a general condition is derived.

Bayesian updating; sensitivity analysis; multiattribute utility functions; discrete probability distributions

nije evidentirano