Let (Ω,F,P) be a probability space, let T⊆F be a sub-σ-algebra, let (R,R,ρ) be a σ-finite measure space, and let W:Ω→R be measurable with respect to F and R. Write R⊗T for the product σ-algebra on R×Ω; measurability of [0,∞]-valued maps and their integrals are those of Lebesgue Integral of a Nonnegative Measurable Function, pointwise products of [0,∞]-valued maps being formed with the multiplication conventions (0⋅∞=0, a⋅∞=∞ for a>0) of Image Measures, Measures with Densities, and Change of Variables, and E denotes the integral with respect to P of a [0,∞]-valued F-measurable map, extending the expectation of a nonnegative random variable. Let f:R×Ω→[0,∞) be measurable with respect to R⊗T and suppose that for every C∈T and every A∈R,
E[1C1A(W)]=E[1C∫R1A(r)f(r,⋅)ρ(dr)]in [0,∞],
where 1S denotes the indicator of a set S and where, for each ω, the inner integral is that of the R-measurable section r↦1A(r)f(r,ω), the map ω↦∫R1A(r)f(r,ω)ρ(dr) being T-measurable by the Tonelli theorem applied on R×Ω with the restriction of P to T. (This hypothesis is the special case Z=1C, g=1A of the identity E[Zg(W)]=E[Z∫Rg(r)f(r,⋅)ρ(dr)] for arbitrary T-measurable Z:Ω→[0,∞] and R-measurable g:R→[0,∞], so it holds whenever that identity does.)
Then for every R⊗T-measurable H:R×Ω→[0,∞]: the map ω↦H(W(ω),ω) is F-measurable; the map ω↦∫RH(r,ω)f(r,ω)ρ(dr) is T-measurable, each inner integral being that of an R-measurable section; and
E[H(W,⋅)]=E[∫RH(r,⋅)f(r,⋅)ρ(dr)]in [0,∞].