Lemma A-1, p. App-1 — the GPD tail Γ_α(β(v − s)) has constant α-virtual value (1 − α)s − 1/β
ProvedSinglePointPricing.Det.lemma_A_1generalized-paretop2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1virtual-value
Fix , and , and consider the generalized Pareto tail for , with density . Its α-virtual value is constant:
The generalized Pareto tails are therefore on the boundary of the α-strongly regular class; this is what makes every element of the worst-case family admissible.
Differences from the source:
- reading: 'The cumulative distribution function Γ_α(β(v − s))' → Γ_α(β(v − s)) is treated as a ccdf, with density f = −d/dv Γ_α(β(v − s)); reason: Γ_α decreases from 1, the α-virtual value is defined with F̄/f (App-1), and the proof differentiates it as a ccdf; 'cumulative distribution function' is a slip
- encoding: α implicit (Γ_α is defined for α ∈ [0, 1], p. 13) → hypothesis α ∈ [0, 1]; reason: the range on which Γ_α is defined
Preamble
import Mathlib import Definitions.Def_SinglePointPricing_Det_Setting
Formal statement
namespace SinglePointPricing.Det
/-- Lemma A-1, p. App-1: for `β > 0` and `s ≥ 0`, the generalized Pareto tail `v ↦ Γ_α(β(v − s))`
on `v ≥ s` has constant `α`-virtual value `(1 − α)v − F̄(v)/f(v) = (1 − α)s − 1/β`, where the density
`f` is minus the derivative of the tail. -/
theorem lemma_A_1 (α β s : ℝ) (hα : α ∈ Set.Icc (0 : ℝ) 1) (hβ : 0 < β) (hs : 0 ≤ s) :
∀ v : ℝ, s ≤ v →
(1 - α) * v - GammaA α (β * (v - s)) / (-deriv (fun x => GammaA α (β * (x - s))) v)
= (1 - α) * s - 1 / β := by sorry
end SinglePointPricing.Det
Source
Allouah, Bahamou & Besbes, Optimal Pricing with a Single Point, arXiv:2103.05611v2, Lemma A-1, p. App-1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.