Step (d), p.10 —
ProvedSuperPareto.step_d_sum_identityheavy-tailsnegative-dependencep2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1probabilityrisk-managementstochastic-dominance
Let , let be events, for , and let be an event with . Then for every , and
Together with display (11) applied to the events and weights , this gives , the last step of the proof of Theorem 1(ii).
Formalization Note Probabilities are real (P.real); the subsets of are the Finsets of a finite index type.
Differences from the source:
- reading: the middle sum is printed ∑ⱼ θᵢ P(Aⱼ) → ∑ⱼ θⱼ P(Aⱼ); reason: the index of θ does not match the summation index; the preceding sum ∑ⱼ θⱼ ∑_{S∋j} P(B_S) fixes the meaning
- generalization: events A₁, …, Aₙ independent of (X₁, …, Xₙ) and an event A independent of X with P(A) = ∑θᵢP(Aᵢ) → any measurable sets Aᵢ and any set A₀ with P(A₀) = ∑θᵢP(Aᵢ), no random variable present; reason: the identity involves only the probabilities of the events; the paper's case is a special case
- generalization: n ≥ 2 random variables on an atomless probability space (standing assumptions, p.4) → a finite index type ι of any size (n = 1 allowed) on any probability space; reason: the paper's case n ≥ 2 on an atomless space is a special case; for n = 1 the statement holds trivially
- encoding: random variables and events → measurable functions Ω → ℝ and measurable sets; reason: a random variable is a measurable function; Lean states measurability explicitly
Preamble
import Mathlib open MeasureTheory
Formal statement
namespace SuperPareto
/-- **Step (d): the weights of the events `B_S` sum to `P(A)`.** Chen, Embrechts, Wang, *An
unexpected stochastic dominance*, arXiv:2208.08471v4, §2.3, step (d), p.10: `∑_{i∈S} θᵢ ∈ [0, 1]`
for every `S ⊆ [n]`, and
`∑_{S⊆[n]} P(B_S) ∑_{i∈S} θᵢ = ∑_j θⱼ ∑_{S∋j} P(B_S) = ∑_j θⱼ P(Aⱼ) = P(A)`,
where `B_S = (⋂_{i∈S} Aᵢ) ∩ (⋂_{i∈Sᶜ} Aᵢᶜ)` and `P(A) = ∑ θᵢ P(Aᵢ)`.
**Formalization Note.** The page prints the third sum as `∑_j θᵢ P(Aⱼ)`, with mismatched indices;
`∑_j θⱼ P(Aⱼ)` is meant. The index set is a `Fintype` `ι`; `Δₙ` is `stdSimplex ℝ ι`; `S ∋ j` is
`Finset.univ.filter (fun S => j ∈ S)`; events are measurable; probabilities are real (`P.real`).
`A₀` is the paper's `A`; only the value of `P(A)` is used. -/
theorem step_d_sum_identity {Ω ι : Type*} [MeasurableSpace Ω] {P : Measure Ω}
[IsProbabilityMeasure P] [Fintype ι] [DecidableEq ι] (θ : ι → ℝ) (hθ : θ ∈ stdSimplex ℝ ι)
(A : ι → Set Ω) (hA : ∀ i, MeasurableSet (A i)) (A₀ : Set Ω)
(hA₀ : P.real A₀ = ∑ i, θ i * P.real (A i)) :
(∀ S : Finset ι, 0 ≤ ∑ i ∈ S, θ i ∧ ∑ i ∈ S, θ i ≤ 1) ∧
∑ S : Finset ι, P.real ((⋂ i ∈ S, A i) ∩ (⋂ (i) (_ : i ∉ S), (A i)ᶜ)) * (∑ i ∈ S, θ i) =
∑ j, θ j * ∑ S ∈ Finset.univ.filter (fun S : Finset ι => j ∈ S),
P.real ((⋂ i ∈ S, A i) ∩ (⋂ (i) (_ : i ∉ S), (A i)ᶜ)) ∧
∑ j, θ j * ∑ S ∈ Finset.univ.filter (fun S : Finset ι => j ∈ S),
P.real ((⋂ i ∈ S, A i) ∩ (⋂ (i) (_ : i ∉ S), (A i)ᶜ)) =
∑ j, θ j * P.real (A j) ∧
∑ j, θ j * P.real (A j) = P.real A₀ := by sorry
end SuperPareto
Source
Chen, Embrechts, Wang, An unexpected stochastic dominance: Pareto distributions, dependence, and diversification, arXiv:2208.08471v4, §2.3, step (d), p.10, "As ∑i∈S θi ∈ [0, 1] for any S ⊆ [n], and ∑S⊆[n] P(BS) ∑i∈S θi = …"
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.