Posterior Information Process for Parametric Families of Experiments.211 Probability, Networks and Algorithms

摘要

In a filtered statistical experiment a priori and a posteriori probability211u001emeasures are defined on an abstract parametric space. The information in the 211u001eposterior, given the prior, is defined by the usual Kullback-Leibler formula. 211u001eCertain properties of this quantity is investigated in the context of so-called 211u001earithmetic and geometric measures and arithmetic and geometric processes. 211u001eInteresting multiplicative decompositions are presented that involve Hellinger 211u001eprocesses indexed both by prior and by posterior distributions.