B.1, p. 26 — the sets D^j_{n,n} = D^j_n ∩ D_{n,n}, j ∈ [n], partition D_{n,n}
ProvedBootRobust.Perf.Dj_partitionbootstrapnearest-neighborsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1prescriptive-analytics
Let and let be a nested chain of neighbourhoods. Write . Then
- for all in , and
Every bootstrap distribution has a unique smallest neighbourhood holding at least of its observations; this partition is what lets the estimator be written as a maximum of partial estimators (Theorem 1).
Formalization Note The chain abstracts Definition 2: under its discrimination property the neighbourhoods are nested, holds all observations, and is the convention for at .
Preamble
import Mathlib import Definitions.Def_BootRobust_Perf_Setting
Formal statement
namespace BootRobust.Perf
/-- B.1, p. 26: the sets `D^j_{n,n} = D^j_n ∩ D_{n,n}`, `j ∈ [n]`, are pairwise disjoint and
cover `D_{n,n}`. -/
theorem Dj_partition {ι : Type*} [Fintype ι] [DecidableEq ι] (n k : ℕ) (hk : 1 ≤ k)
(hkn : k ≤ n) (N : ℕ → Finset ι) (hNmono : Monotone N) (hN0 : N 0 = ∅)
(hNn : N n = Finset.univ) :
(∀ j ∈ Finset.Icc 1 n, ∀ j' ∈ Finset.Icc 1 n, j ≠ j' →
Disjoint (Dj n k N j ∩ Dnn n) (Dj n k N j' ∩ Dnn n)) ∧
(⋃ j ∈ Finset.Icc 1 n, Dj n k N j ∩ Dnn n) = Dnn n := by sorry
end BootRobust.Perf
Source
Bertsimas and Van Parys, Bootstrap robust prescriptive analytics, arXiv:1711.09974v2, B.1 (proof of Theorem 1), p. 26, "Formally, D^j_{n,n} ∩ D^{j′}_{n,n} = ∅ …"
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.