Work in the probability space (Ω, G, P restricted to G): the indefinite integral of X over the sets of G is a finite signed measure absolutely continuous with respect to it, so the Radon–Nikodym theorem gives a G-measurable density, and integrals of G-measurable functions are the same for P and its restriction. Uniqueness is the uniqueness of densities, and the vector case is componentwise.
Throughout, denotes the restriction of to , a probability measure on (Finite Signed Measures from Finite Measures and from Integrals (iii)); it is -finite, the constant sequence witnessing this (Measure, Measure Space, and Probability Measure (iv)), so the Radon–Nikodym theorem applies in the measure space . A -measurable is -measurable, it lies in if and only if it lies in , and then
(Integrals with Respect to the Restriction of a Measure to a Sub-σ-Algebra (ii), Integrals with Respect to the Restriction of a Measure to a Sub-σ-Algebra (iv), with Integrals over Measurable Subsets (vi) for integrability on ).
Existence. Let . The function on ,
is a finite signed measure on that is absolutely continuous with respect to (Finite Signed Measures from Finite Measures and from Integrals (iv)). Under the axiom of dependent choice, the Radon–Nikodym theorem for finite signed measures, applied in (The Radon–Nikodym Theorem for a σ-Finite Measure, and Uniqueness of Densities (iv)), gives with for every . As an element of , is -measurable (Integrable Functions and Their Lebesgue Integral (ii)), so by the transfer above and
This proves (i).
Uniqueness. Let and be as in (ii). By the transfer, both lie in and
The uniqueness of densities in (The Radon–Nikodym Theorem for a σ-Finite Measure, and Uniqueness of Densities (i)), which uses no choice, shows that belongs to and has -measure , that is, .
The vector case. Let ; its components lie in (Integrable Vector- and Matrix-Valued Functions and Their Integrals (i)). A function is -measurable, respectively -measurable, if and only if all its components are -measurable, respectively -measurable (Vector- and Matrix-Valued Measurable Maps: Components, Entries and Continuous Images (i), applied on and on ).
For uniqueness, let and be as in (iii). For each , (ii) applied to , and shows that lies in with . Since , by induction on this set lies in and (Closure Properties of σ-Algebras, Intersections and Traces of σ-Algebras, and the Generated σ-Algebra (vii), Basic Properties of a Measure (vii)). This proves (iii); no choice is used.
For existence, assume the axiom of dependent choice. For let be the set of -measurable with for every ; it is nonempty by (i). Since is finite (Finite Sets: the Pigeonhole Principle, Uniqueness of the Length, Subsets, Unions, Products, Images, Bounded Sets of Natural Numbers, Extreme Elements, Sets of Maps, Finite Unions and Finite Choice (vii)), finite choice (Finite Sets: the Pigeonhole Principle, Uniqueness of the Length, Subsets, Unions, Products, Images, Bounded Sets of Natural Numbers, Extreme Elements, Sets of Maps, Finite Unions and Finite Choice (xi)), applied to the family of nonempty sets, gives a family with for every . Let be the function with components for (Component Maps of a Map into , Entry Maps of a Map into Matrices, and Maps Assembled from Them (iii)). Its components are -measurable and lie in , so is -measurable, -measurable, and , with for every and . This proves (iv), using no choice beyond (i).
Uses of dependent choice. The existence statements use the axiom of dependent choice only through The Radon–Nikodym Theorem for a σ-Finite Measure, and Uniqueness of Densities (iv); the uniqueness statements use no choice.
Loading…