Lemma 4.1, p. 17 — relaxation of (1): V^LB ≤ V^OPT
ProvedCorrClusterExp.IBR.lemma_4_1correlation-matrixexperimental-designp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1robust-optimization
Let and let be cluster sizes. Let and be the optimal values of the min–max problems (1) (over experiments with marginals ) and (5) (over all correlation matrices). Then
The relaxation (5) is what the paper's approximation analysis compares IBR experiments against.
Differences from the source:
- generalization: sizes sorted, (Remark 4.1) → any sizes ; reason: the order of the clusters plays no role, and the sorted case is an instance.
Preamble
import Mathlib import Definitions.Def_CorrClusterExp_IBR_Setting
Formal statement
namespace CorrClusterExp.IBR
theorem lemma_4_1 (q : ℝ) (hq0 : 0 < q) (hq : q ≤ 1 / 2) (n : ℕ) (w : Fin n → ℝ)
(hw0 : ∀ i, 0 ≤ w i) :
VLB w ≤ VOPT q w := by sorry
end CorrClusterExp.IBR
Source
Candogan, Chen, Niazadeh, Correlated Cluster-Based Randomized Experiments: Robust Variance Minimization, SSRN 3852100 (version of March 22, 2023), p. 17, Lemma 4.1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.