Theorem 3.11 — on spaces of strong negative type, dcov(θ) = 0 implies θ is a product measure
ProvedDistCov.Indep.theorem_3_11Let and be separable metric spaces with their Borel -fields, both of strong negative type, and let be a probability measure on whose marginals on and on have finite first moments. If
then is a product measure: .
This is the key result of the theory: on spaces of strong negative type, distance covariance vanishes only under independence, which makes the distance-covariance test of independence consistent against all alternatives.
Formalization Note Strong negative type is defined from (3.1) and the condition " only when " (p. 11), not through an embedding. is the integral of against ; the finite-first-moment hypotheses on the marginals make it a genuine integral. "Product measure" means equality with the product of the two marginals of . Separability (SecondCountableTopology) and the Borel -fields are the standing assumption added by the author in Errata (i), p. 24.
import Mathlib import Definitions.Def_DistCov_Indep_Setting
namespace DistCov.Indep
open MeasureTheory
theorem theorem_3_11 {X : Type*} [MetricSpace X] [MeasurableSpace X] [BorelSpace X] [SecondCountableTopology X] {Y : Type*} [MetricSpace Y] [MeasurableSpace Y] [BorelSpace Y] [SecondCountableTopology Y]
(hX : StrongNegType X) (hY : StrongNegType Y)
(θ : Measure (X × Y)) [IsProbabilityMeasure θ]
(hμ : FiniteFirstMoment (θ.map Prod.fst)) (hν : FiniteFirstMoment (θ.map Prod.snd))
(h : dcov θ = 0) : IsProduct θ := by sorry
end DistCov.Indep
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.