feat(probability/kernel/cond_distrib): regular conditional probability distributions (#19090)

We define the regular conditional probability distribution `cond_distrib Y X μ` of `Y : α → Ω` given `X : α → β`, where `Ω` is a standard Borel space. This is a `kernel β Ω` such that for almost all `a`, for all measurable set `s`, `cond_distrib Y X μ (X a) s` is equal to the conditional expectation `μ⟦Y ⁻¹' s | mβ.comap X⟧` evaluated at `a`.

Also define the above notation for the conditional expectation of the indicator of a set.



Co-authored-by: Rémy Degenne <remydegenne@gmail.com>

Author: RemyDegenne

src/measure_theory/constructions/prod/basic.lean

@@ -749,6 +749,21 @@ instance [is_finite_measure ρ] : is_finite_measure ρ.fst := by { rw fst, apply
 instance [is_probability_measure ρ] : is_probability_measure ρ.fst :=
 { measure_univ := by { rw fst_univ, exact measure_univ, } }
 
+lemma fst_map_prod_mk₀ {X : α → β} {Y : α → γ} {μ : measure α}
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ) :
+  (μ.map (λ a, (X a, Y a))).fst = μ.map X :=
+begin
+  ext1 s hs,
+  rw [measure.fst_apply hs, measure.map_apply_of_ae_measurable (hX.prod_mk hY) (measurable_fst hs),
+    measure.map_apply_of_ae_measurable hX hs, ← prod_univ, mk_preimage_prod, preimage_univ,
+    inter_univ],
+end
+
+lemma fst_map_prod_mk {X : α → β} {Y : α → γ} {μ : measure α}
+  (hX : measurable X) (hY : measurable Y) :
+  (μ.map (λ a, (X a, Y a))).fst = μ.map X :=
+fst_map_prod_mk₀ hX.ae_measurable hY.ae_measurable
+
 /-- Marginal measure on `β` obtained from a measure on `ρ` `α × β`, defined by `ρ.map prod.snd`. -/
 noncomputable
 def snd (ρ : measure (α × β)) : measure β := ρ.map prod.snd
@@ -764,6 +779,21 @@ instance [is_finite_measure ρ] : is_finite_measure ρ.snd := by { rw snd, apply
 instance [is_probability_measure ρ] : is_probability_measure ρ.snd :=
 { measure_univ := by { rw snd_univ, exact measure_univ, } }
 
+lemma snd_map_prod_mk₀ {X : α → β} {Y : α → γ} {μ : measure α}
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ) :
+  (μ.map (λ a, (X a, Y a))).snd = μ.map Y :=
+begin
+  ext1 s hs,
+  rw [measure.snd_apply hs, measure.map_apply_of_ae_measurable (hX.prod_mk hY) (measurable_snd hs),
+    measure.map_apply_of_ae_measurable hY hs, ← univ_prod, mk_preimage_prod, preimage_univ,
+    univ_inter],
+end
+
+lemma snd_map_prod_mk {X : α → β} {Y : α → γ} {μ : measure α}
+  (hX : measurable X) (hY : measurable Y) :
+  (μ.map (λ a, (X a, Y a))).snd = μ.map Y :=
+snd_map_prod_mk₀ hX.ae_measurable hY.ae_measurable
+
 end measure
 
 

src/measure_theory/function/conditional_expectation/basic.lean

@@ -187,6 +187,14 @@ lemma ae_eq_trim_iff_of_ae_strongly_measurable' {α β} [topological_space β] [
 ⟨λ h, hfm.ae_eq_mk.trans (h.trans hgm.ae_eq_mk.symm),
   λ h, hfm.ae_eq_mk.symm.trans (h.trans hgm.ae_eq_mk)⟩
 
+lemma ae_strongly_measurable.comp_ae_measurable'
+  {α β γ : Type*} [topological_space β] {mα : measurable_space α} {mγ : measurable_space γ}
+  {f : α → β} {μ : measure γ} {g : γ → α}
+  (hf : ae_strongly_measurable f (μ.map g)) (hg : ae_measurable g μ) :
+  ae_strongly_measurable' (mα.comap g) (f ∘ g) μ :=
+⟨(hf.mk f) ∘ g, hf.strongly_measurable_mk.comp_measurable (measurable_iff_comap_le.mpr le_rfl),
+  ae_eq_comp hg hf.ae_eq_mk⟩
+
 /-- If the restriction to a set `s` of a σ-algebra `m` is included in the restriction to `s` of
 another σ-algebra `m₂` (hypothesis `hs`), the set `s` is `m` measurable and a function `f` almost
 everywhere supported on `s` is `m`-ae-strongly-measurable, then `f` is also

src/probability/kernel/cond_distrib.lean

@@ -0,0 +1,315 @@
+/-
+Copyright (c) 2023 Rémy Degenne. All rights reserved.
+Released under Apache 2.0 license as described in the file LICENSE.
+Authors: Rémy Degenne
+-/
+import probability.kernel.disintegration
+import probability.notation
+
+/-!
+# Regular conditional probability distribution
+
+We define the regular conditional probability distribution of `Y : α → Ω` given `X : α → β`, where
+`Ω` is a standard Borel space. This is a `kernel β Ω` such that for almost all `a`, `cond_distrib`
+evaluated at `X a` and a measurable set `s` is equal to the conditional expectation
+`μ⟦Y ⁻¹' s | mβ.comap X⟧` evaluated at `a`.
+
+`μ⟦Y ⁻¹' s | mβ.comap X⟧` maps a measurable set `s` to a function `α → ℝ≥0∞`, and for all `s` that
+map is unique up to a `μ`-null set. For all `a`, the map from sets to `ℝ≥0∞` that we obtain that way
+verifies some of the properties of a measure, but in general the fact that the `μ`-null set depends
+on `s` can prevent us from finding versions of the conditional expectation that combine into a true
+measure. The standard Borel space assumption on `Ω` allows us to do so.
+
+The case `Y = X = id` is developed in more detail in `probability/kernel/condexp.lean`: here `X` is
+understood as a map from `Ω` with a sub-σ-algebra to `Ω` with its default σ-algebra and the
+conditional distribution defines a kernel associated with the conditional expectation with respect
+to `m`.
+
+## Main definitions
+
+* `cond_distrib Y X μ`: regular conditional probability distribution of `Y : α → Ω` given
+  `X : α → β`, where `Ω` is a standard Borel space.
+
+## Main statements
+
+* `cond_distrib_ae_eq_condexp`: for almost all `a`, `cond_distrib` evaluated at `X a` and a
+  measurable set `s` is equal to the conditional expectation `μ⟦Y ⁻¹' s | mβ.comap X⟧ a`.
+* `condexp_prod_ae_eq_integral_cond_distrib`: the conditional expectation
+  `μ[(λ a, f (X a, Y a)) | X ; mβ]` is almost everywhere equal to the integral
+  `∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a))`.
+
+-/
+
+open measure_theory set filter topological_space
+
+open_locale ennreal measure_theory probability_theory
+
+namespace probability_theory
+
+variables {α β Ω F : Type*}
+  [topological_space Ω] [measurable_space Ω] [polish_space Ω] [borel_space Ω] [nonempty Ω]
+  [normed_add_comm_group F]
+  {mα : measurable_space α} {μ : measure α} [is_finite_measure μ] {X : α → β} {Y : α → Ω}
+
+/-- **Regular conditional probability distribution**: kernel associated with the conditional
+expectation of `Y` given `X`.
+For almost all `a`, `cond_distrib Y X μ` evaluated at `X a` and a measurable set `s` is equal to
+the conditional expectation `μ⟦Y ⁻¹' s | mβ.comap X⟧ a`. It also satisfies the equality
+`μ[(λ a, f (X a, Y a)) | mβ.comap X] =ᵐ[μ] λ a, ∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a))` for
+all integrable functions `f`. -/
+@[irreducible] noncomputable
+def cond_distrib {mα : measurable_space α} [measurable_space β]
+  (Y : α → Ω) (X : α → β) (μ : measure α) [is_finite_measure μ] :
+  kernel β Ω :=
+(μ.map (λ a, (X a, Y a))).cond_kernel
+
+instance [measurable_space β] : is_markov_kernel (cond_distrib Y X μ) :=
+by { rw cond_distrib, apply_instance, }
+
+variables {mβ : measurable_space β} {s : set Ω} {t : set β} {f : β × Ω → F}
+include mβ
+
+section measurability
+
+lemma measurable_cond_distrib (hs : measurable_set s) :
+  measurable[mβ.comap X] (λ a, cond_distrib Y X μ (X a) s) :=
+(kernel.measurable_coe _ hs).comp (measurable.of_comap_le le_rfl)
+
+lemma _root_.measure_theory.ae_strongly_measurable.ae_integrable_cond_distrib_map_iff
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf : ae_strongly_measurable f (μ.map (λ a, (X a, Y a)))) :
+  (∀ᵐ a ∂(μ.map X), integrable (λ ω, f (a, ω)) (cond_distrib Y X μ a))
+    ∧ integrable (λ a, ∫ ω, ‖f (a, ω)‖ ∂(cond_distrib Y X μ a)) (μ.map X)
+  ↔ integrable f (μ.map (λ a, (X a, Y a))) :=
+by rw [cond_distrib, ← hf.ae_integrable_cond_kernel_iff, measure.fst_map_prod_mk₀ hX hY]
+
+variables [normed_space ℝ F] [complete_space F]
+
+lemma _root_.measure_theory.ae_strongly_measurable.integral_cond_distrib_map
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf : ae_strongly_measurable f (μ.map (λ a, (X a, Y a)))) :
+  ae_strongly_measurable (λ x, ∫ y, f (x, y) ∂(cond_distrib Y X μ x)) (μ.map X) :=
+by { rw [← measure.fst_map_prod_mk₀ hX hY, cond_distrib], exact hf.integral_cond_kernel, }
+
+lemma _root_.measure_theory.ae_strongly_measurable.integral_cond_distrib
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf : ae_strongly_measurable f (μ.map (λ a, (X a, Y a)))) :
+  ae_strongly_measurable (λ a, ∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a))) μ :=
+(hf.integral_cond_distrib_map hX hY).comp_ae_measurable hX
+
+lemma ae_strongly_measurable'_integral_cond_distrib
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf : ae_strongly_measurable f (μ.map (λ a, (X a, Y a)))) :
+  ae_strongly_measurable' (mβ.comap X) (λ a, ∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a))) μ :=
+(hf.integral_cond_distrib_map hX hY).comp_ae_measurable' hX
+
+end measurability
+
+section integrability
+
+lemma integrable_to_real_cond_distrib (hX : ae_measurable X μ) (hs : measurable_set s) :
+  integrable (λ a, (cond_distrib Y X μ (X a) s).to_real) μ :=
+begin
+  refine integrable_to_real_of_lintegral_ne_top _ _,
+  { exact measurable.comp_ae_measurable (kernel.measurable_coe _ hs) hX, },
+  { refine ne_of_lt _,
+    calc ∫⁻ a, cond_distrib Y X μ (X a) s ∂μ
+        ≤ ∫⁻ a, 1 ∂μ : lintegral_mono (λ a, prob_le_one)
+    ... = μ univ : lintegral_one
+    ... < ∞ : measure_lt_top _ _, },
+end
+
+lemma _root_.measure_theory.integrable.cond_distrib_ae_map
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  ∀ᵐ b ∂(μ.map X), integrable (λ ω, f (b, ω)) (cond_distrib Y X μ b) :=
+by { rw [cond_distrib, ← measure.fst_map_prod_mk₀ hX hY], exact hf_int.cond_kernel_ae, }
+
+lemma _root_.measure_theory.integrable.cond_distrib_ae
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  ∀ᵐ a ∂μ, integrable (λ ω, f (X a, ω)) (cond_distrib Y X μ (X a)) :=
+ae_of_ae_map hX (hf_int.cond_distrib_ae_map hX hY)
+
+lemma _root_.measure_theory.integrable.integral_norm_cond_distrib_map
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  integrable (λ x, ∫ y, ‖f (x, y)‖ ∂(cond_distrib Y X μ x)) (μ.map X) :=
+by { rw [cond_distrib, ← measure.fst_map_prod_mk₀ hX hY], exact hf_int.integral_norm_cond_kernel, }
+
+lemma _root_.measure_theory.integrable.integral_norm_cond_distrib
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  integrable (λ a, ∫ y, ‖f (X a, y)‖ ∂(cond_distrib Y X μ (X a))) μ :=
+(hf_int.integral_norm_cond_distrib_map hX hY).comp_ae_measurable hX
+
+variables [normed_space ℝ F] [complete_space F]
+
+lemma _root_.measure_theory.integrable.norm_integral_cond_distrib_map
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  integrable (λ x, ‖∫ y, f (x, y) ∂(cond_distrib Y X μ x)‖) (μ.map X) :=
+by { rw [cond_distrib, ← measure.fst_map_prod_mk₀ hX hY], exact hf_int.norm_integral_cond_kernel, }
+
+lemma _root_.measure_theory.integrable.norm_integral_cond_distrib
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  integrable (λ a, ‖∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a))‖) μ :=
+(hf_int.norm_integral_cond_distrib_map hX hY).comp_ae_measurable hX
+
+lemma _root_.measure_theory.integrable.integral_cond_distrib_map
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  integrable (λ x, ∫ y, f (x, y) ∂(cond_distrib Y X μ x)) (μ.map X) :=
+(integrable_norm_iff (hf_int.1.integral_cond_distrib_map hX hY)).mp
+  (hf_int.norm_integral_cond_distrib_map hX hY)
+
+lemma _root_.measure_theory.integrable.integral_cond_distrib
+  (hX : ae_measurable X μ) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  integrable (λ a, ∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a))) μ :=
+(hf_int.integral_cond_distrib_map hX hY).comp_ae_measurable hX
+
+end integrability
+
+lemma set_lintegral_preimage_cond_distrib (hX : measurable X) (hY : ae_measurable Y μ)
+  (hs : measurable_set s) (ht : measurable_set t) :
+  ∫⁻ a in X ⁻¹' t, cond_distrib Y X μ (X a) s ∂μ = μ (X ⁻¹' t ∩ Y ⁻¹' s) :=
+by rw [lintegral_comp (kernel.measurable_coe _ hs) hX, cond_distrib,
+  ← measure.restrict_map hX ht, ← measure.fst_map_prod_mk₀ hX.ae_measurable hY,
+  set_lintegral_cond_kernel_eq_measure_prod _ ht hs,
+  measure.map_apply_of_ae_measurable (hX.ae_measurable.prod_mk hY) (ht.prod hs),
+  mk_preimage_prod]
+
+lemma set_lintegral_cond_distrib_of_measurable_set (hX : measurable X) (hY : ae_measurable Y μ)
+  (hs : measurable_set s) {t : set α} (ht : measurable_set[mβ.comap X] t) :
+  ∫⁻ a in t, cond_distrib Y X μ (X a) s ∂μ = μ (t ∩ Y ⁻¹' s) :=
+by { obtain ⟨t', ht', rfl⟩ := ht, rw set_lintegral_preimage_cond_distrib hX hY hs ht', }
+
+/-- For almost every `a : α`, the `cond_distrib Y X μ` kernel applied to `X a` and a measurable set
+`s` is equal to the conditional expectation of the indicator of `Y ⁻¹' s`. -/
+lemma cond_distrib_ae_eq_condexp (hX : measurable X) (hY : measurable Y) (hs : measurable_set s) :
+  (λ a, (cond_distrib Y X μ (X a) s).to_real) =ᵐ[μ] μ⟦Y ⁻¹' s | mβ.comap X⟧ :=
+begin
+  refine ae_eq_condexp_of_forall_set_integral_eq hX.comap_le _ _ _ _,
+  { exact (integrable_const _).indicator (hY hs),  },
+  { exact λ t ht _, (integrable_to_real_cond_distrib hX.ae_measurable hs).integrable_on, },
+  { intros t ht _,
+    rw [integral_to_real ((measurable_cond_distrib hs).mono hX.comap_le le_rfl).ae_measurable
+        (eventually_of_forall (λ ω, measure_lt_top (cond_distrib Y X μ (X ω)) _)),
+      integral_indicator_const _ (hY hs), measure.restrict_apply (hY hs), smul_eq_mul, mul_one,
+      inter_comm, set_lintegral_cond_distrib_of_measurable_set hX hY.ae_measurable hs ht], },
+  { refine (measurable.strongly_measurable _).ae_strongly_measurable',
+    exact @measurable.ennreal_to_real _ (mβ.comap X) _ (measurable_cond_distrib hs), },
+end
+
+/-- The conditional expectation of a function `f` of the product `(X, Y)` is almost everywhere equal
+to the integral of `y ↦ f(X, y)` against the `cond_distrib` kernel. -/
+lemma condexp_prod_ae_eq_integral_cond_distrib' [normed_space ℝ F] [complete_space F]
+  (hX : measurable X) (hY : ae_measurable Y μ)
+  (hf_int : integrable f (μ.map (λ a, (X a, Y a)))) :
+  μ[(λ a, f (X a, Y a)) | mβ.comap X] =ᵐ[μ] λ a, ∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a)) :=
+begin
+  have hf_int' : integrable (λ a, f (X a, Y a)) μ,
+  { exact (integrable_map_measure hf_int.1 (hX.ae_measurable.prod_mk hY)).mp hf_int, },
+  refine (ae_eq_condexp_of_forall_set_integral_eq hX.comap_le hf_int' (λ s hs hμs, _) _ _).symm,
+  { exact (hf_int.integral_cond_distrib hX.ae_measurable hY).integrable_on, },
+  { rintros s ⟨t, ht, rfl⟩ _,
+    change ∫ a in X ⁻¹' t, ((λ x', ∫ y, f (x', y) ∂(cond_distrib Y X μ) x') ∘ X) a ∂μ
+      = ∫ a in X ⁻¹' t, f (X a, Y a) ∂μ,
+    rw ← integral_map hX.ae_measurable,
+    swap,
+    { rw ← measure.restrict_map hX ht,
+      exact (hf_int.1.integral_cond_distrib_map hX.ae_measurable hY).restrict, },
+    rw [← measure.restrict_map hX ht, ← measure.fst_map_prod_mk₀ hX.ae_measurable hY, cond_distrib,
+      set_integral_cond_kernel_univ_right ht hf_int.integrable_on,
+      set_integral_map (ht.prod measurable_set.univ) hf_int.1 (hX.ae_measurable.prod_mk hY),
+      mk_preimage_prod, preimage_univ, inter_univ], },
+  { exact ae_strongly_measurable'_integral_cond_distrib hX.ae_measurable hY hf_int.1, },
+end
+
+/-- The conditional expectation of a function `f` of the product `(X, Y)` is almost everywhere equal
+to the integral of `y ↦ f(X, y)` against the `cond_distrib` kernel. -/
+lemma condexp_prod_ae_eq_integral_cond_distrib₀ [normed_space ℝ F] [complete_space F]
+  (hX : measurable X) (hY : ae_measurable Y μ)
+  (hf : ae_strongly_measurable f (μ.map (λ a, (X a, Y a))))
+  (hf_int : integrable (λ a, f (X a, Y a)) μ) :
+  μ[(λ a, f (X a, Y a)) | mβ.comap X] =ᵐ[μ] λ a, ∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a)) :=
+begin
+  have hf_int' : integrable f (μ.map (λ a, (X a, Y a))),
+  { rwa integrable_map_measure hf (hX.ae_measurable.prod_mk hY), },
+  exact condexp_prod_ae_eq_integral_cond_distrib' hX hY hf_int',
+end
+
+/-- The conditional expectation of a function `f` of the product `(X, Y)` is almost everywhere equal
+to the integral of `y ↦ f(X, y)` against the `cond_distrib` kernel. -/
+lemma condexp_prod_ae_eq_integral_cond_distrib [normed_space ℝ F] [complete_space F]
+  (hX : measurable X) (hY : ae_measurable Y μ)
+  (hf : strongly_measurable f) (hf_int : integrable (λ a, f (X a, Y a)) μ) :
+  μ[(λ a, f (X a, Y a)) | mβ.comap X] =ᵐ[μ] λ a, ∫ y, f (X a, y) ∂(cond_distrib Y X μ (X a)) :=
+begin
+  have hf_int' : integrable f (μ.map (λ a, (X a, Y a))),
+  { rwa integrable_map_measure hf.ae_strongly_measurable (hX.ae_measurable.prod_mk hY), },
+  exact condexp_prod_ae_eq_integral_cond_distrib' hX hY hf_int',
+end
+
+lemma condexp_ae_eq_integral_cond_distrib [normed_space ℝ F] [complete_space F]
+  (hX : measurable X) (hY : ae_measurable Y μ)
+  {f : Ω → F} (hf : strongly_measurable f) (hf_int : integrable (λ a, f (Y a)) μ) :
+  μ[(λ a, f (Y a)) | mβ.comap X] =ᵐ[μ] λ a, ∫ y, f y ∂(cond_distrib Y X μ (X a)) :=
+condexp_prod_ae_eq_integral_cond_distrib hX hY (hf.comp_measurable measurable_snd) hf_int
+
+/-- The conditional expectation of `Y` given `X` is almost everywhere equal to the integral
+`∫ y, y ∂(cond_distrib Y X μ (X a))`. -/
+lemma condexp_ae_eq_integral_cond_distrib' {Ω} [normed_add_comm_group Ω] [normed_space ℝ Ω]
+  [complete_space Ω] [measurable_space Ω] [borel_space Ω] [second_countable_topology Ω] {Y : α → Ω}
+  (hX : measurable X) (hY_int : integrable Y μ) :
+  μ[Y | mβ.comap X] =ᵐ[μ] λ a, ∫ y, y ∂(cond_distrib Y X μ (X a)) :=
+condexp_ae_eq_integral_cond_distrib hX hY_int.1.ae_measurable strongly_measurable_id hY_int
+
+lemma _root_.measure_theory.ae_strongly_measurable.comp_snd_map_prod_mk
+  {Ω F} {mΩ : measurable_space Ω} {X : Ω → β} {μ : measure Ω}
+  [topological_space F] (hX : measurable X) {f : Ω → F} (hf : ae_strongly_measurable f μ) :
+  ae_strongly_measurable (λ x : β × Ω, f x.2) (μ.map (λ ω, (X ω, ω))) :=
+begin
+  refine ⟨λ x, hf.mk f x.2, hf.strongly_measurable_mk.comp_measurable measurable_snd, _⟩,
+  suffices h : measure.quasi_measure_preserving prod.snd (μ.map (λ ω, (X ω, ω))) μ,
+  { exact measure.quasi_measure_preserving.ae_eq h hf.ae_eq_mk, },
+  refine ⟨measurable_snd, measure.absolutely_continuous.mk (λ s hs hμs, _)⟩,
+  rw measure.map_apply _ hs,
+  swap, { exact measurable_snd, },
+  rw measure.map_apply,
+  { rw [← univ_prod, mk_preimage_prod, preimage_univ, univ_inter, preimage_id'],
+    exact hμs, },
+  { exact hX.prod_mk measurable_id, },
+  { exact measurable_snd hs, },
+end
+
+lemma _root_.measure_theory.integrable.comp_snd_map_prod_mk {Ω} {mΩ : measurable_space Ω}
+  {X : Ω → β} {μ : measure Ω} (hX : measurable X) {f : Ω → F} (hf_int : integrable f μ) :
+  integrable (λ x : β × Ω, f x.2) (μ.map (λ ω, (X ω, ω))) :=
+begin
+  have hf := hf_int.1.comp_snd_map_prod_mk hX,
+  refine ⟨hf, _⟩,
+  rw [has_finite_integral, lintegral_map' hf.ennnorm (hX.prod_mk measurable_id).ae_measurable],
+  exact hf_int.2,
+end
+
+lemma ae_strongly_measurable_comp_snd_map_prod_mk_iff {Ω F} {mΩ : measurable_space Ω}
+  [topological_space F] {X : Ω → β} {μ : measure Ω} (hX : measurable X) {f : Ω → F} :
+  ae_strongly_measurable (λ x : β × Ω, f x.2) (μ.map (λ ω, (X ω, ω)))
+    ↔ ae_strongly_measurable f μ :=
+⟨λ h, h.comp_measurable (hX.prod_mk measurable_id), λ h, h.comp_snd_map_prod_mk hX⟩
+
+lemma integrable_comp_snd_map_prod_mk_iff {Ω} {mΩ : measurable_space Ω} {X : Ω → β} {μ : measure Ω}
+  (hX : measurable X) {f : Ω → F} :
+  integrable (λ x : β × Ω, f x.2) (μ.map (λ ω, (X ω, ω))) ↔ integrable f μ :=
+⟨λ h, h.comp_measurable (hX.prod_mk measurable_id), λ h, h.comp_snd_map_prod_mk hX⟩
+
+lemma condexp_ae_eq_integral_cond_distrib_id [normed_space ℝ F] [complete_space F]
+  {X : Ω → β} {μ : measure Ω} [is_finite_measure μ]
+  (hX : measurable X) {f : Ω → F} (hf_int : integrable f μ) :
+  μ[f | mβ.comap X] =ᵐ[μ] λ a, ∫ y, f y ∂(cond_distrib id X μ (X a)) :=
+condexp_prod_ae_eq_integral_cond_distrib' hX ae_measurable_id (hf_int.comp_snd_map_prod_mk hX)
+
+end probability_theory