Documentation

Atlas.TheoryOfProbability.code.MartingaleOrthogonality

On a finite measure space, square-integrability (together with a.e. strong measurability) implies integrability. This follows from the elementary bound |Y| ≤ Y^2 + 1.

theorem integrable_mul_of_sq_integrable {Ω : Type u_1} {m0 : MeasurableSpace Ω} {μ : MeasureTheory.Measure Ω} {f g : Ω} (hf : MeasureTheory.Integrable f μ) (hg : MeasureTheory.Integrable g μ) (hf2 : MeasureTheory.Integrable (fun (ω : Ω) => f ω ^ 2) μ) (hg2 : MeasureTheory.Integrable (fun (ω : Ω) => g ω ^ 2) μ) :
MeasureTheory.Integrable (fun (ω : Ω) => f ω * g ω) μ

If f and g are both integrable and square-integrable, then their pointwise product ω ↦ f ω * g ω is integrable. This is the Cauchy–Schwarz-style bound |fg| ≤ (f² + g²)/2.

theorem martingale_orthogonal_general {Ω : Type u_1} {m0 : MeasurableSpace Ω} {μ : MeasureTheory.Measure Ω} [MeasureTheory.IsProbabilityMeasure μ] { : MeasureTheory.Filtration m0} {X : Ω} (hmart : MeasureTheory.Martingale X μ) {m n : } (hmn : m n) {Y : Ω} (hY_meas : MeasureTheory.StronglyMeasurable Y) (hY_int : MeasureTheory.Integrable (fun (ω : Ω) => Y ω ^ 2) μ) (hX_sq : ∀ (k : ), MeasureTheory.Integrable (fun (ω : Ω) => X k ω ^ 2) μ) :
(ω : Ω), (X n ω - X m ω) * Y ω μ = 0

Orthogonality of martingale increments (Lecture 28): if X is a martingale with E X_n^2 < ∞ for all n, and m ≤ n, then for any ℱ m-measurable random variable Y with E Y^2 < ∞ we have E[(X_n - X_m) · Y] = 0. In other words the increment X_n - X_m is orthogonal in L^2 to every ℱ_m-measurable random variable.