Aldous's uniform modulus estimate #
First bound the number of genuine exits in probability. Then exclude short gaps among a fixed finite number of exits and control a deterministic terminal window. Merging the last block produces a partition admissible for Billingsley's modulus.
theorem
BGW.Skorokhod.exists_uniform_aldous_modulus_bound
{T : ℝ}
(hT : 0 < T)
{Ω : ℕ → Type u_1}
[m : (n : ℕ) → MeasurableSpace (Ω n)]
(P : (n : ℕ) → MeasureTheory.Measure (Ω n))
[∀ (n : ℕ), MeasureTheory.IsProbabilityMeasure (P n)]
(𝓕 : (n : ℕ) → MeasureTheory.Filtration (↑(Set.Icc 0 T)) (m n))
(X : (n : ℕ) → Ω n → D T ℝ)
(hX : ∀ (n : ℕ), Measurable (X n))
(hadapted : ∀ (n : ℕ) (t : ↑(Set.Icc 0 T)), Measurable fun (ω : Ω n) => (X n ω) t)
(hAldous :
∀ (τ : (n : ℕ) → Ω n → ↑(Set.Icc 0 T)),
(∀ (n : ℕ), IsStopTime (𝓕 n) (τ n)) →
∀ (h : ℕ → ℝ),
(∀ (n : ℕ), 0 ≤ h n) →
Filter.Tendsto h Filter.atTop (nhds 0) →
∀ ε > 0,
Filter.Tendsto
(fun (n : ℕ) => (P n) {ω : Ω n | ε ≤ |(X n ω) (Set.projIcc 0 T ⋯ (↑(τ n ω) + h n)) - (X n ω) (τ n ω)|})
Filter.atTop (nhds 0))
{ε η : ℝ}
(hε : 0 < ε)
(hη : 0 < η)
:
Aldous's sequence condition makes the modulus small in probability for all large indices.
Reference: Billingsley 1999, Section 16.