Documentation

Linglib.Core.MeasureTheory.Measure.StochasticOrder

The stochastic order on measures #

On a preordered measurable space, μ ≤ₛₜ ν says that ν gives every measurable upper set at least the mass μ does: ν sits higher in the order. The relation is preserved by pushing forward along a monotone map, and it is what makes the integral of a monotone function monotone in the measure.

The order is usually established through a coupling: two random variables on one probability space with the given laws, the first almost surely below the second (stochasticLE_map_of_ae_le). Couplings of the coordinates assemble into a coupling of the product, so the order passes to product measures.

TODO #

Strassen's theorem, that on a Polish space the stochastic order yields a monotone coupling, is not stated.

def MeasureTheory.Measure.StochasticLE {α : Type u_1} [MeasurableSpace α] [Preorder α] (μ ν : Measure α) :

The stochastic order: ν gives every measurable upper set at least the mass μ does.

Equations
  • μ.StochasticLE ν = ∀ ⦃U : Set α⦄, IsUpperSet U → MeasurableSet U → μ U ≤ ν U
Instances For
    def MeasureTheory.Measure.«term_≤ₛₜ_» :
    Lean.TrailingParserDescr

    The stochastic order: ν gives every measurable upper set at least the mass μ does.

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      theorem MeasureTheory.Measure.StochasticLE.refl {α : Type u_1} [MeasurableSpace α] [Preorder α] (μ : Measure α) :
      theorem MeasureTheory.Measure.StochasticLE.trans {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν ξ : Measure α} (h₁ : μ.StochasticLE ν) (h₂ : ν.StochasticLE ξ) :
      instance MeasureTheory.Measure.StochasticLE.instRefl {α : Type u_1} [MeasurableSpace α] [Preorder α] :
      Std.Refl fun (x1 x2 : Measure α) => x1.StochasticLE x2
      instance MeasureTheory.Measure.StochasticLE.instIsTrans {α : Type u_1} [MeasurableSpace α] [Preorder α] :
      IsTrans (Measure α) fun (x1 x2 : Measure α) => x1.StochasticLE x2
      theorem MeasureTheory.Measure.StochasticLE.map {α : Type u_1} {β : Type u_2} [MeasurableSpace α] [MeasurableSpace β] [Preorder α] [Preorder β] {μ ν : Measure α} (h : μ.StochasticLE ν) {f : α → β} (hf : Monotone f) (hfm : Measurable f) :
      (map f μ).StochasticLE (map f ν)

      Pushing forward along a monotone map preserves the stochastic order.

      theorem MeasureTheory.Measure.StochasticLE.map_of_ae_eq {α : Type u_1} {β : Type u_2} [MeasurableSpace α] [MeasurableSpace β] [Preorder α] [Preorder β] {μ ν : Measure α} (h : μ.StochasticLE ν) {f g : α → β} (hg : Monotone g) (hgm : Measurable g) (hμ : f =ᵐ[μ] g) (hν : f =ᵐ[ν] g) :
      (map f μ).StochasticLE (map f ν)

      A map that agrees almost everywhere with a monotone one preserves the stochastic order.

      theorem MeasureTheory.Measure.StochasticLE.measure_le_of_isLowerSet {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν : Measure α} [IsFiniteMeasure μ] (h : μ.StochasticLE ν) (huniv : μ Set.univ = ν Set.univ) {L : Set α} (hL : IsLowerSet L) (hLm : MeasurableSet L) :
      ν L ≤ μ L

      Measures of equal finite mass in the stochastic order compare on lower sets the other way.

      theorem MeasureTheory.Measure.StochasticLE.lintegral_ofReal_le {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν : Measure α} (h : μ.StochasticLE ν) {f : α → ℝ} (hf : Monotone f) (hfm : Measurable f) :
      ∫⁻ (a : α), ENNReal.ofReal (f a) ∂μ ≤ ∫⁻ (a : α), ENNReal.ofReal (f a) ∂ν

      The lower integral of the positive part of a monotone function is monotone in the measure.

      theorem MeasureTheory.Measure.StochasticLE.lintegral_ofReal_le_of_antitone {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν : Measure α} [IsFiniteMeasure μ] (h : μ.StochasticLE ν) (huniv : μ Set.univ = ν Set.univ) {f : α → ℝ} (hf : Antitone f) (hfm : Measurable f) :
      ∫⁻ (a : α), ENNReal.ofReal (f a) ∂ν ≤ ∫⁻ (a : α), ENNReal.ofReal (f a) ∂μ

      The lower integral of the positive part of an antitone function is antitone in the measure, between measures of equal finite mass.

      theorem MeasureTheory.Measure.StochasticLE.integral_le {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν : Measure α} [IsFiniteMeasure μ] (h : μ.StochasticLE ν) (huniv : μ Set.univ = ν Set.univ) {f : α → ℝ} (hf : Monotone f) (hfm : Measurable f) (hμ : Integrable f μ) (hν : Integrable f ν) :
      ∫ (a : α), f a ∂μ ≤ ∫ (a : α), f a ∂ν

      The integral of a monotone function is monotone in the measure, between measures of equal finite mass.

      theorem MeasureTheory.Measure.StochasticLE.integral_le_of_antitone {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν : Measure α} [IsFiniteMeasure μ] (h : μ.StochasticLE ν) (huniv : μ Set.univ = ν Set.univ) {f : α → ℝ} (hf : Antitone f) (hfm : Measurable f) (hμ : Integrable f μ) (hν : Integrable f ν) :
      ∫ (a : α), f a ∂ν ≤ ∫ (a : α), f a ∂μ

      The integral of an antitone function is antitone in the measure, between measures of equal finite mass.

      theorem MeasureTheory.Measure.stochasticLE_map_of_ae_le {α : Type u_1} {Ω : Type u_3} [MeasurableSpace α] {mΩ : MeasurableSpace Ω} [Preorder α] {P : Measure Ω} {X Y : Ω → α} (hX : Measurable X) (hY : Measurable Y) (h : X ≤ᵐ[P] Y) :
      (map X P).StochasticLE (map Y P)

      Two random variables on one space, the first almost surely below the second, have stochastically ordered laws.

      structure MeasureTheory.Measure.IsMonotoneCoupling {α : Type u_1} [MeasurableSpace α] [Preorder α] (γ : Measure (α × α)) (μ ν : Measure α) :

      A monotone coupling of μ and ν: a measure on pairs with marginals μ and ν under which the first coordinate is almost surely below the second.

      • map_fst : map Prod.fst γ = μ
      • map_snd : map Prod.snd γ = ν
      • ae_le : ∀ᵐ (z : α × α) ∂γ, z.1 ≤ z.2
      Instances For
        theorem MeasureTheory.Measure.IsMonotoneCoupling.stochasticLE {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν : Measure α} {γ : Measure (α × α)} (h : γ.IsMonotoneCoupling μ ν) :

        Measures with a monotone coupling are stochastically ordered.

        theorem MeasureTheory.Measure.IsMonotoneCoupling.isProbabilityMeasure {α : Type u_1} [MeasurableSpace α] [Preorder α] {μ ν : Measure α} {γ : Measure (α × α)} [IsProbabilityMeasure μ] (h : γ.IsMonotoneCoupling μ ν) :
        IsProbabilityMeasure γ
        theorem MeasureTheory.Measure.stochasticLE_infinitePi {ι : Type u_4} [Countable ι] {X : ι → Type u_5} [(i : ι) → MeasurableSpace (X i)] [(i : ι) → Preorder (X i)] {μ ν : (i : ι) → Measure (X i)} [∀ (i : ι), IsProbabilityMeasure (μ i)] {γ : (i : ι) → Measure (X i × X i)} (h : ∀ (i : ι), (γ i).IsMonotoneCoupling (μ i) (ν i)) :
        (infinitePi μ).StochasticLE (infinitePi ν)

        Monotone couplings of the coordinates give the stochastic order between countable products.