The role of coherence for handling probabilistic evaluations and independence (Q2576604)

The author handles independence statements and conditional probability assessments in a coherent setting. When a probability is not complete (i.e. it is defined on an arbitrary set of events) the independence statements are not automatically induced by the probability values. So it is necessarry to distinguish between the statements that are induced and those are compatible. Given a set of independence statements it is shown how to find its maximal subset of statements induced by a conditional probability assessment. Moreover, the problem of compatibility between qualitative knowledge (independence statements) and numerical knowledge (probability values) is studied, and a procedure based on linear optimization is proposed.