Lemma EC.5 — Poisson tail with a bounded random rate
ProvedMarkdownReadjust.lemma_EC_5dynamic-pricingmarkdown-pricingp2o-gran-per-chapterp2o-plan-paperp2o-v1poisson-processrevenue-management
Let and let be a random variable. Let be a random variable that, conditionally on , is Poisson with parameter . Then for every ,
This concentration bound is what makes the bad event unlikely (Lemma EC.4).
Differences from the source:
- encoding: random variables Λ ∈ (0,λ] and X with X | Λ ~ Poisson(Λ) → an arbitrary probability measure μ on ℝ with μ((0,λ]ᶜ) = 0 and P(X − Λ ≥ ε) = ∫ Po(Λ){n : n − Λ ≥ ε} dμ(Λ); reason: the joint law of (Λ, X) is this mixture; same instances
Preamble
import Mathlib open MeasureTheory ProbabilityTheory
Formal statement
namespace MarkdownReadjust
/-- **Lemma EC.5, p.ec24** (Chen–Jasin, SSRN 3740871). Let `Λ ∈ (0, λ]` be a random variable
and let `X`, conditionally on `Λ`, be Poisson with parameter `Λ`. Then for every `ε > 0`,
`P(X − Λ ≥ ε) ≤ exp(−0.1 ε min{ε/λ, 1})`.
**Formalization Note.** The law of `Λ` is an arbitrary probability measure `μ` on `ℝ`
concentrated on `(0, λ]`, and `P(X − Λ ≥ ε) = ∫ Po(Λ){n : n − Λ ≥ ε} dμ(Λ)` (the mixture). -/
theorem lemma_EC_5 (lam : ℝ) (hlam : 0 < lam) (μ : Measure ℝ) [IsProbabilityMeasure μ]
(hμ : μ (Set.Ioc 0 lam)ᶜ = 0) (ε : ℝ) (hε : 0 < ε) :
∫⁻ Λ, poissonMeasure Λ.toNNReal {n : ℕ | ε ≤ (n : ℝ) - Λ} ∂μ
≤ ENNReal.ofReal (Real.exp (-0.1 * ε * min (ε / lam) 1)) := by sorry
end MarkdownReadjust
Source
Chen, Jasin, Simple monotonic readjustment policies with applications to markdown pricing and pricing in the presence of strategic customers, SSRN 3740871 (2023-10-09), Lemma EC.5, p.ec24
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.