Posterior Integration of Independent Stochastic Estimates

Kryazhimskiy, A.V. (2013). Posterior Integration of Independent Stochastic Estimates. IIASA Interim Report. IIASA, Laxenburg, Austria: IR-13-006

Preview

Text IR-13-006.pdf Download (413kB) | Preview

Abstract

We develop a unified approach to posterior integration of prior stochastic estimates (probability distributions) provided by independent statistically inaccurate observation methods. Our departure point is the posterior event formed in the product of the probability spaces associated with the prior stochastic estimates. The Bayesian probability conditioned to the posterior event has identical projections onto the coordinate spaces; its common projection is defined to be the posterior integrated stochastic estimate. We view integration as a binary operation on the set of all probabilities on a given finite set of elementary events and analyze its algebraic properties. We show how integration changes the information quality of the integrated probabilities and study integral convergence properties of infinite sequences of probabilities.

Actions (login required)

View Item