본문으로 바로 이동 / Skip to content

PROOF 164

Prove.

The definition of conditional probability shrinks the sample space

On a probability space.

A way to read this proof

First separate what is assumed from what is to be proved. At each step, ask which definition or earlier result permits the move. If it is unclear, return to that premise rather than skipping to the conclusion.

Definition D1

When P(B) > 0, the probability Q under the information that B has occurred is defined as Q(A) = P(A ∩ B) / P(B), written P(A|B).

Proof.

(1)If B has occurred, no outcome outside B has occurred. The new whole space is B, not S.convention
(2)Therefore what remains of an event A is only A ∩ B, and Q(A) is determined by P(A∩B).(1)
(3)Setting Q(A) = P(A∩B) gives Q(S) = P(S∩B) = P(B), which in general is not 1.(2), A7, E
(4)A probability must equal 1 on the whole space. So the scale is reset to P(B).(3), A2
(5)Define Q(A) = P(A∩B) / P(B). This definition applies only to B with P(B) > 0.(4), A4

∎

Lemma L1

If P(B) = 0, then for every event A, P(A ∩ B) = 0

Proof.

(1)B = (A∩B) ∪ (B∖A), and the two events are disjoint.A11
(2)P(B) = P(A∩B) + P(B∖A)(1), A3
(3)P(A∩B) ≥ 0 and P(B∖A) ≥ 0A1
(4)Suppose P(B) = 0.assumption
(5)Both terms are non-negative and their sum is 0, so P(A∩B) = 0(2), (3), (4), A12
(6)Then both numerator and denominator of D1 become 0, and by Proof 02 the value is not uniquely determined. This is why the proviso P(B) > 0 is there.(5), D1, A4

∎

Theorem T1

If P(B) > 0, then Q satisfies all three of Kolmogorov's axioms.

Proof.

(1)Suppose P(B) > 0.assumption
(2)P(A∩B) ≥ 0A1
(3)Q(A) = P(A∩B)/P(B) ≥ 0. This is P1.(1), (2), D1, A5
(4)S ∩ B = BA7
(5)Q(S) = P(S∩B)/P(B) = P(B)/P(B) = 1. This is P2.(1), (4), D1, A10, E
(6)Suppose X ∩ Y = ∅.assumption
(7)(X ∪ Y) ∩ B = (X∩B) ∪ (Y∩B)A8
(8)X∩B and Y∩B are disjoint.(6), A9
(9)P((X∪Y)∩B) = P(X∩B) + P(Y∩B)(7), (8), A3
(10)Q(X ∪ Y) = [P(X∩B) + P(Y∩B)] / P(B)(9), D1
(11)= P(X∩B)/P(B) + P(Y∩B)/P(B) = Q(X) + Q(Y). This is P3.(10), A6, D1
(12)Q satisfies all three axioms.(3), (5), (11)

∎

Corollary C1

Every proposition proved from the three axioms alone holds for P(A|B) as well.

Proof.

(1)Q satisfies the three axioms.T1
(2)Therefore any proposition resting only on the three axioms applies to Q.(1)
(3)By Proof 161, 0 ≤ P(A|B) ≤ 1(2)
(4)By Proof 162, P(Aᶜ|B) = 1 − P(A|B)(2)
(5)Therefore calling P(A∩B)/P(B) a probability is not a name but a credential.(1), D1
(6)The P(B) in the denominator is not a convention. It is the scale that makes Q(S) = 1, and without that scale the credential in (5) does not stand.(5), D1, A2

∎

Remark

Can A4 be removed from this proof?

P(B) > 0 looks like a proviso attached to the definition. Remove it and see what the definition loses.