(5.8), p. 544 — for independent trials, Eh(W) = 𝒫_λh − Σ_i p_i² E U_λh(W^{(i)})
ProvedPoissonDepTrials.SecondOrder.identity_5_8Let be independent Bernoulli trials with , , and . For every bounded on the nonnegative integers,
This is the first-order identity (2.6) specialised to independent trials (); applying it a second time to produces the second-order term of Theorem 5.1.
Formalization Note The page writes with a random index uniform on and independent of the trials, and ; this equals , the form stated here (the random indices are, in the paper's words, a device so that "the use of symbols may be simplified"). The trials are a sequence with almost surely and for and , the paper's own padding convention (p. 535); constants generate the trivial -algebra, so independence of the padded family is independence of . is added because divides by . is replaced by a bound .
import Mathlib import Definitions.Def_PoissonDepTrials_SecondOrder_Setting open MeasureTheory ProbabilityTheory Finset
namespace PoissonDepTrials.SecondOrder
/-- (5.8), proof of Theorem 5.1, p. 544, in index form: for independent Bernoulli trials,
`Eh(W) = 𝒫_λh − Σ_{i=1}^n p_i² E U_λh(W^{(i)})`. -/
theorem identity_5_8 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(n : ℕ) (X : ℕ → Ω → ℕ) (hX : IsBernoulliTrials P n X) (hind : iIndepFun (fun i => X i) P)
(hlam : 0 < lam P n X) (h : ℕ → ℝ) (M : ℝ) (hM : ∀ k, |h k| ≤ M) :
∫ ω, h (W n X ω) ∂P = poissonExp (lam P n X) h -
∑ i ∈ Finset.Icc 1 n, prob P X i ^ 2 * ∫ ω, stU (lam P n X) h (Wi n X i ω) ∂P := by sorry
end PoissonDepTrials.SecondOrder
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.