Dempster's rule for combining two belief functions assumes the independence of the sources of information. If this assumption is questionable, I suggest to use the least specific combination minimizing the conflict among the ones allowed by a simple generalization of Dempster's rule. This increases the monotonicity of the reasoning and helps us to manage situations of dependence. Some properties of this combination rule and its usefulness in a generalization of Bayes' theorem are then considered.
Keywords. belief functions, propositional logic, combination, Dempster's rule, independence, conflict, monotonicity, nonspecificity, idempotency, associativity, Bayes' theorem
Authors addresses:
Seminar für Statistik, LEO C13
ETH Zentrum
CH-8092 Zürich
E-mail addresses:
Marco Cattaneo | cattaneo@stat.math.ethz.ch |