Conditioning and Martingales
Let (Ω,F, P) be a probability space and A, B ∈F events such that P(B) > 0. In this section we seek to formalize the intuitive content of the phrase “probability of A given (or, equivalently, “conditioned on”) B”. The words “given B ” imply prior knowledge of the fact that the sample point ω lies in B. It is then natural to consider the reduced probability space (Ω, Fb, Pb) with ΩB = B,FB = A∩B A ∈ F (the “trace σ-field”), PB(C) = P(C)/P(B) for C ∈ TB, and define the probability of A given B, denoted by P(A/B), as PB(A ∩ B) = P(AB)/P(B). The reader may convince himself of the reasonableness of this procedure by evaluating, say, the probability of two consecutive heads in four tosses of a fair coin given the knowledge that the number of heads was even, using the principle of insufficient reason.
KeywordsConditional Expectation Dominate Convergence Theorem Polish Space Martingale Theory Real Random Variable
Unable to display preview. Download preview PDF.