Lemma D.1 (convex-quantile contraction): if Y has convex quantile Q and Y^[t] has quantile Q(tu), then NV(Y^[t])≤t·NV(Y) and NV(Z+Y^[t])≤(1-t)NV(Z)+t·NV(Z+Y)≤NV(Z+Y)
ProvedOWMR.convex_quantile_contractioninventoryinventory-theoryoperations-researchrenewal-theorystochastic-control
Let be a probability space, real random variables, nondecreasing and convex on , and . Suppose is integrable with quantile function , that is, has the law of for uniform on ; has quantile function on ; and is integrable and independent of and of . With as in (D.1) (the paper's minimum, written as an infimum),
The quantile function may be unbounded; only integrability of and is assumed.
Preamble
import Mathlib.Probability.Independence.Basic import Mathlib.MeasureTheory.Integral.Bochner.Basic import Mathlib.MeasureTheory.Measure.Lebesgue.Basic import Mathlib.Analysis.Convex.Function open MeasureTheory ProbabilityTheory open scoped ENNReal NNReal
Formal statement
namespace OWMR
theorem convex_quantile_contraction {Ω : Type} [MeasurableSpace Ω] (P : Measure Ω)
[IsProbabilityMeasure P] (Q : ℝ → ℝ) (hQm : MonotoneOn Q (Set.Ioo 0 1))
(hQc : ConvexOn ℝ (Set.Ioo 0 1) Q) (Y Yt Z : Ω → ℝ) (hY : Measurable Y)
(hYt : Measurable Yt) (hZ : Measurable Z) (hYi : Integrable Y P) (hZi : Integrable Z P)
(t a b : ℝ) (ht0 : 0 < t) (ht1 : t ≤ 1) (ha : 0 < a) (hb : 0 < b)
(hYlaw : P.map Y = (volume.restrict (Set.Ioo 0 1)).map Q)
(hYtlaw : P.map Yt = (volume.restrict (Set.Ioo 0 1)).map (fun u => Q (t * u)))
(hZY : IndepFun Z Y P) (hZYt : IndepFun Z Yt P) :
let NV : (Ω → ℝ) → ℝ := fun X =>
⨅ z : ℝ, ∫ ω, (a * max (z - X ω) 0 + b * max (X ω - z) 0) ∂P
NV Yt ≤ t * NV Y ∧
NV (fun ω => Z ω + Yt ω) ≤ (1 - t) * NV Z + t * NV (fun ω => Z ω + Y ω) ∧
(1 - t) * NV Z + t * NV (fun ω => Z ω + Y ω) ≤ NV (fun ω => Z ω + Y ω) := by sorry
end OWMRSource
M. Hu, "A 5/2 Guarantee for Echelon Stock (s,S) Policies in One-Warehouse Multi-Retailer Systems", working paper (manuscript of November 2026), Appendix D, Lemma D.1 (Convex-Quantile Contraction) with the optimized loss (D.1) and inequalities (D.2) and (D.3), p. E.24; proof in Appendix D.1, p. E.27.
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.