Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 1 — the plug-in stochastic kriging predictor (13) is unbiased under Assumption 1

Proved
StochKriging.Unbiased.theorem_1

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

krigingp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1simulationunbiasedness

Consider the stochastic kriging model Yj(x)=β0+M(x)+εj(x)\mathcal Y_j(\mathbf x)=\beta_0+\mathsf M(\mathbf x)+\varepsilon_j(\mathbf x)Yj​(x)=β0​+M(x)+εj​(x) on Rd\mathbb R^dRd, with distinct design points x1,…,xk\mathbf x_1,\dots,\mathbf x_kx1​,…,xk​, replication counts ni≥2n_i\ge2ni​≥2 and noise variances V(xi)>0\mathsf V(\mathbf x_i)>0V(xi​)>0. Let Yˉ\bar{\mathcal Y}Yˉ​ be the vector of sample means, V^(xi)=S2(xi)\widehat{\mathsf V}(\mathbf x_i)=\mathcal S^2(\mathbf x_i)V(xi​)=S2(xi​) the sample variances of the same replications,

Σ^ε=Diag{V^(x1)/n1,…,V^(xk)/nk},\widehat\Sigma_\varepsilon=\mathrm{Diag}\big\{\widehat{\mathsf V}(\mathbf x_1)/n_1,\dots,\widehat{\mathsf V}(\mathbf x_k)/n_k\big\},Σε​=Diag{V(x1​)/n1​,…,V(xk​)/nk​},

and define, for any point x0∈Rd\mathbf x_0\in\mathbb R^dx0​∈Rd (simulated or not),

Y^^(x0)=β0+ΣM(x0,⋅)⊤[ΣM+Σ^ε]−1(Yˉ−β01k).(13)\widehat{\widehat{\mathsf Y}}(\mathbf x_0)=\beta_0+\Sigma_{\mathsf M}(\mathbf x_0,\cdot)^\top\big[\Sigma_{\mathsf M}+\widehat\Sigma_\varepsilon\big]^{-1}\big(\bar{\mathcal Y}-\beta_0\mathbf 1_k\big).\tag{13}Y(x0​)=β0​+ΣM​(x0​,⋅)⊤[ΣM​+Σε​]−1(Yˉ​−β0​1k​).(13)

If Assumption 1 holds, then the prediction error Y^^(x0)−Y(x0)\widehat{\widehat{\mathsf Y}}(\mathbf x_0)-\mathsf Y(\mathbf x_0)Y(x0​)−Y(x0​), where Y(x0)=β0+M(x0)\mathsf Y(\mathbf x_0)=\beta_0+\mathsf M(\mathbf x_0)Y(x0​)=β0​+M(x0​), is integrable and

E[Y^^(x0)−Y(x0)]=0.\mathrm E\Big[\widehat{\widehat{\mathsf Y}}(\mathbf x_0)-\mathsf Y(\mathbf x_0)\Big]=0.E[Y(x0​)−Y(x0​)]=0.

Estimating the intrinsic variance from the replications, and plugging it into the stochastic kriging predictor, therefore introduces no prediction bias; the price of estimation is paid only in mean squared error.

Formalization Note Integrability of the error is stated explicitly, because a Bochner integral of a non-integrable function is 000 in Lean. ΣM\Sigma_{\mathsf M}ΣM​ and ΣM(x0,⋅)\Sigma_{\mathsf M}(\mathbf x_0,\cdot)ΣM​(x0​,⋅) are the true covariances of the field; the inverse is Matrix.inv, which is a genuine inverse here because Assumption 1 makes ΣM\Sigma_{\mathsf M}ΣM​ positive definite and Σ^ε\widehat\Sigma_\varepsilonΣε​ is positive semidefinite. x0\mathbf x_0x0​ may coincide with a design point.

Preamble
import Mathlib
import Definitions.Def_StochKriging_Unbiased_Model
import Definitions.Def_StochKriging_Unbiased_Predictor

open MeasureTheory ProbabilityTheory
open scoped NNReal
Formal statement
namespace StochKriging.Unbiased

/-- Theorem 1 (p. 366): under Assumption 1, the plug-in stochastic kriging predictor (13) is
unbiased for `Y(x₀) = β₀ + M(x₀)`: `E[Ŷ̂(x₀) − Y(x₀)] = 0` (the error is integrable). -/
theorem theorem_1 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
    {d k : ℕ} (x : Fin k → EuclideanSpace ℝ (Fin d)) (hx : Function.Injective x)
    (n : Fin k → ℕ) (hn : ∀ i, 2 ≤ n i) (β₀ : ℝ) (V : EuclideanSpace ℝ (Fin d) → ℝ≥0)
    (hV : ∀ i, 0 < V (x i)) (M : EuclideanSpace ℝ (Fin d) → Ω → ℝ)
    (ε : ℕ → EuclideanSpace ℝ (Fin d) → Ω → ℝ) (hA : Assumption1 P M ε V x)
    (x₀ : EuclideanSpace ℝ (Fin d)) :
    Integrable (fun ω => predictorHat P β₀ M ε x n x₀ ω - (β₀ + M x₀ ω)) P ∧
      ∫ ω, (predictorHat P β₀ M ε x n x₀ ω - (β₀ + M x₀ ω)) ∂P = 0 := by sorry

end StochKriging.Unbiased
Source
Ankenman, Nelson, Staum, Stochastic Kriging for Simulation Metamodeling, Proc. 2008 Winter Simulation Conference, p. 366, Theorem 1
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me