Proposition 1, pairwise form, p. 13 — the w-biased experiment beats the n-biased one iff p > p̂
ProvedExpBias.OwnBias.proposition_1_pairwisebiased-experimentexperimentationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1value-of-information
Assume the standing assumptions. Comparing only the two biased experiments and , each valued at the firm's best signal-contingent decision rule,
This is the choice "between the w-biased and n-biased experiment" that Proposition 1 describes: a firm that would exploit without experimenting prefers the experiment whose only decisive signal is bad news. The goal of the mission strengthens it to optimality over the whole experiment set.
Formalization Note. The inequalities are strict; strictness uses and (that is, ).
Preamble
import Mathlib import Definitions.Def_ExpBias_OwnBias_Model
Formal statement
namespace ExpBias.OwnBias
theorem proposition_1_pairwise (P : Params) (hP : P.Standing) :
(pHat P < P.p → value P (nBiased P) < value P (wBiased P)) ∧
(P.p < pHat P → value P (wBiased P) < value P (nBiased P)) := by sorry
end ExpBias.OwnBias
Source
Gans, A Synthetic Model of Disruption and Experimentation, NBER Working Paper 29091 (July 2021), p. 13 (PDF 14), §2.2, Proposition 1 and the paragraph after it; proof on p. 31 (PDF 32)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.