Lemma 2, p. 38 — for X_n ∼ Poisson(λn), Z = sup_{n ≥ 1} X_n/n has finite mean
ProvedMarketInterference.FluidLimit.lemma_2integrabilityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1poisson-distribution
Let and, on a probability space, let be random variables with for every ; no independence or other joint structure is assumed. Put . Then
In the proof of Theorem 1 this supplies an integrable dominating function for the scaled matching values.
Formalization Note The paper writes ; the maximum need not be attained, so is the supremum, taken in , and the conclusion is that its lower Lebesgue integral is finite. Each is a measurable -valued map whose law is .
Preamble
import Mathlib
Formal statement
namespace MarketInterference.FluidLimit
open MeasureTheory ProbabilityTheory NNReal ENNReal
/-- Lemma 2 (arXiv:2205.02274v4, p. 38): if `X_n ∼ Poisson(λ n)` for every `n ≥ 1` (any joint law)
and `Y_n = X_n / n`, then `Z = sup_{n ≥ 1} Y_n` has a finite mean. -/
theorem lemma_2 {Ω : Type} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(lam : ℝ≥0) (X : ℕ → Ω → ℕ)
(hX : ∀ n : ℕ, 1 ≤ n → Measurable (X n) ∧ P.map (X n) = poissonMeasure ((n : ℝ≥0) * lam)) :
∫⁻ ω, (⨆ (n : ℕ) (_ : 1 ≤ n), ENNReal.ofReal ((X n ω : ℝ) / (n : ℝ))) ∂P < ⊤ := by sorry
end MarketInterference.FluidLimit
Source
Bright, Delarue and Lobel, Reducing Marketplace Interference Bias Via Shadow Prices, arXiv:2205.02274v4, p. 38, Lemma 2
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.