p. 16, after Definition 3 — every run of a pigeonhole design puts T/2 subjects in each group
ProvedPigeonholeDesign.Upper.assign_balancedbalanceexperimental-designp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1pigeonhole-design
Let be even. For every partition of the covariate space into pigeonholes, every arrival sequence , and every realization of the fair coins, the pigeonhole design assigns exactly subjects to the control group and exactly subjects to the treated group:
This is the role of Step 1 of Definition 3. It guarantees that the discrepancy of every run is a genuine perfect matching between groups of equal size.
Formalization Note The statement holds for an arbitrary covariate type and an arbitrary labelling of pigeonholes. It is the guard against the junk value that the formal discrepancy takes on unbalanced splits: it shows that the designs never produce one.
Preamble
import Mathlib import Definitions.Def_PigeonholeDesign_Upper_Setting
Formal statement
namespace PigeonholeDesign.Upper
theorem assign_balanced {E ι : Type*} [DecidableEq ι] {T : ℕ} (hT : Even T)
(hole : E → ι) (x : Fin T → E) (coin : Fin T → Bool) :
(Finset.univ.filter (fun t => assign hole x coin t = false)).card = T / 2 ∧
(Finset.univ.filter (fun t => assign hole x coin t = true)).card = T / 2 := by sorry
end PigeonholeDesign.Upper
Source
Zhao and Zhou, Pigeonhole Design: Balancing Sequential Experiments from an Online Matching Perspective, arXiv:2201.12936v6, p. 16, remark following Definition 3
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.