Benefit Probability Bounds
Abstract
The two Boolean potential-outcome marginals give algebraic bounds on the benefit mass of every normalized nonnegative joint law.
Theorem 1.1 (Potential-outcome marginals bound the benefit probability).
Proof. Machine-checked in Lean as D5/S3/ConceptDynamics/Causal/BenefitProbabilityBounds.benefit_probability_bounds (✓ std3). ∎
Source. Repository-derived.
Commentary.
Let mass be a normalized nonnegative joint law of the Boolean pair of potential outcomes. The benefit probability is the mass of the false-true response type.
The treatment-one marginal is the sum of the false-true and true-true masses. The treatment-zero marginal is the sum of the true-false and true-true masses.
Nonnegativity of the true-false cell gives the lower marginal-difference bound. Nonnegativity of the true-true and false-false cells gives the two upper bounds.
References
- Truth anchor:
D5/S3/ConceptDynamics/Causal/BenefitProbabilityBounds.benefit_probability_bounds - Dependency: D5/S3/ConceptDynamics/Causal/PrincipalStrata