(8), p. 53 — for t ≥ L, x_t is N(x₀, σ²Σ_{i<L}(1 + iα)²), and Σ_{i<L}(1+iα)² = L(1 + α(L−1) + α²(L−1)(2L−1)/6)
ProvedGravesIMA.Stages.eq_8Throughout, time is indexed by the integers, , and is the single-stage system of Graves (1999, §2) driven by a shock sequence : demand , for ; the exponentially weighted moving-average forecast , ; orders for and for with lead time ; inventory for , with the initial inventory free.
The shocks , , are random variables on a probability space ; they are measurable, mutually independent, and each has law . The system equations hold for every outcome , and the initial inventory is a deterministic constant.
Then for every period , the inventory is normally distributed with
The second equality is stated as the identity of the quantities under the square roots, ; both sides are nonnegative, so it is equivalent to the equality of the standard deviations. The standard deviation of the inventory is the safety stock requirement of the system.
Formalization Note Independence is iIndepFun of the family indexed by the subtype ; shocks at are never read by the model. "Normally distributed with mean and standard deviation " is the law equality P.map X = gaussianReal m (s²); the variance is passed as Real.toNNReal of a sum of squares times , which is nonnegative, so no clipping occurs. Each is assumed measurable, so the pushforward measures are genuine laws.
import Mathlib import Definitions.Def_GravesIMA_Stages_Model open MeasureTheory ProbabilityTheory
namespace GravesIMA.Stages
theorem eq_8 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(σ : ℝ) (ε : ℤ → Ω → ℝ) (hmeas : ∀ t, Measurable (ε t))
(hind : iIndepFun (fun t : {s : ℤ // 1 ≤ s} => ε t) P)
(hlaw : ∀ t : ℤ, 1 ≤ t → P.map (ε t) = gaussianReal 0 (Real.toNNReal (σ ^ 2)))
(μ α : ℝ) (hα0 : 0 ≤ α) (hα1 : α ≤ 1) (L : ℕ) (d F q x : ℤ → Ω → ℝ)
(hsys : ∀ ω, IsStage μ α L (fun s => ε s ω) (fun s => d s ω) (fun s => F s ω)
(fun s => q s ω) (fun s => x s ω))
(x0 : ℝ) (hx0 : ∀ ω, x 0 ω = x0) (t : ℤ) (ht1 : 1 ≤ t) (htL : (L : ℤ) ≤ t) :
P.map (x t) =
gaussianReal x0 (Real.toNNReal (σ ^ 2 * ∑ i ∈ Finset.range L, (1 + (i : ℝ) * α) ^ 2)) ∧
∑ i ∈ Finset.range L, (1 + (i : ℝ) * α) ^ 2 =
(L : ℝ) * (1 + α * ((L : ℝ) - 1) + α ^ 2 * ((L : ℝ) - 1) * (2 * (L : ℝ) - 1) / 6) := by sorry
end GravesIMA.Stages
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.