Proposition 3(a), p. 1348 (σ_ε = 0) — for ω < 1 the optimal forecast q*_QF is strictly positive, finite, and the unique solution of (3)
ProvedTsayQF.EffQF.proposition_3aLet , , be the cost data and the transfer price. Let have law with finite variance and a differentiable, strictly increasing distribution function ; with demand equals . Take a QF contract with and . The retailer's forecast objective is
where the EM builds . Write for if and if . Assume
Then equation (3),
has exactly one solution , and a forecast maximizes over if and only if and solves (3). In particular the optimal forecast exists, is strictly positive and finite, and is the unique solution of (3).
The left side of (3) is the retailer's expected marginal benefit of raising the forecast (more guaranteed availability), the right side its expected marginal cost (a larger minimum purchase); the condition is the QF analogue of the newsvendor fractile.
Formalization Note. This is the instance of the printed statement, whose integration regions read and . At , is piecewise linear and is taken as its right derivative; its value on the -null diagonal does not affect (3), which then reads . The displayed hypothesis on is the explicit form of the paper's presumption that is almost certainly nonnegative (under which and the hypothesis holds); it is exactly what makes the optimal forecast strictly positive. The EM's production is fixed at , as §6.1 does after Proposition 2. The price range generalizes the standing , because the mission applies the result at .
import Mathlib import Definitions.Def_TsayQF_EffQF_Model open MeasureTheory ProbabilityTheory
namespace TsayQF.EffQF
/-- Proposition 3(a), p. 1348, at `σ_ε = 0`: when `ω < 1`, the retailer's optimal forecast
`q*_QF` is strictly positive and finite, and is the unique solution of (3). -/
theorem proposition_3a (D : Data) (ν : Measure ℝ) [IsProbabilityMeasure ν]
(hΘ : StrictMono (cdf ν)) (hΘd : Differentiable ℝ (cdf ν)) (hν : MemLp id 2 ν)
(c : ℝ) (hcu : D.u < c) (hcp : c < D.p + D.s)
(α ω : ℝ) (hω0 : 0 ≤ ω) (hω1 : ω < 1) (hα : -ω ≤ α)
(h0 : (1 - ω) * (c - D.u) * cdf ν 0 < (1 + α) * (D.p + D.s - c) * (1 - cdf ν 0)) :
let eq3 : ℝ → Prop := fun q =>
(1 + α) * ∫ μ in Set.Ici (q * (1 + α)), Gderiv D c (q * (1 + α)) μ ∂ν =
-((1 - ω) * ∫ μ in Set.Iic (q * (1 - ω)), Gderiv D c (q * (1 - ω)) μ ∂ν)
(∃! q, 0 < q ∧ eq3 q) ∧
∀ q, 0 ≤ q → (IsMaxOn (forecastProfit D ν c α ω) (Set.Ici 0) q ↔ (0 < q ∧ eq3 q)) := by sorry
end TsayQF.EffQF
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.