Lemma 9.4 — ∫₀^∞ |P(X > t) − P(Y > t)| t^{κ−1} dt ≤ κ⁻¹E|(X⁺)^κ − (Y⁺)^κ|, and (9.18) when finite
ProvedGoldieRenewal.Kesten.lemma_9_4heavy-tailsmomentsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1tail-integrals
Let and let , be real random variables on a common probability space; .
- (9.17), as an inequality in :
- (9.18): if , then is integrable on and
The lemma converts the tail-difference integrals of the implicit renewal theorem into moment expressions; it is what turns Theorem 2.3 into Corollary 2.4.
Formalization Note The paper prints (9.17) as an equality "finite or infinite". Equality fails in general (for , independent and identically distributed and non-degenerate the left side is while the right side is positive); the paper's proof establishes, and Corollary 2.4 uses, only the inequality, which is what is stated. Both sides of (9.17) are lower Lebesgue integrals, so infinite values are allowed.
Preamble
import Mathlib
Formal statement
namespace GoldieRenewal.Kesten
open MeasureTheory ProbabilityTheory
open scoped ENNReal
/-- **Lemma 9.4** (Goldie, *Implicit renewal theory and tails of solutions of random equations*,
Ann. Appl. Probab. 1(1) (1991), p. 149). Let `κ > 0` and let `X`, `Y` be random variables on a
common probability space.
* (9.17), corrected to an inequality:
`∫₀^∞ |P(X > t) − P(Y > t)| t^{κ−1} dt ≤ κ⁻¹ E|(X⁺)^κ − (Y⁺)^κ|`, finite or infinite.
* (9.18): when `E|(X⁺)^κ − (Y⁺)^κ| < ∞`, the function `t ↦ (P(X > t) − P(Y > t)) t^{κ−1}` is
integrable on `(0, ∞)` and `∫₀^∞ (P(X > t) − P(Y > t)) t^{κ−1} dt = κ⁻¹ E((X⁺)^κ − (Y⁺)^κ)`.
**Formalization Note** (9.17) is printed as an equality; only `≤` holds (for `X`, `Y` i.i.d. and
non-degenerate the left side is `0` and the right side positive), and the paper's proof and its
use in Corollary 2.4 need only `≤`. Both sides of (9.17) are lower Lebesgue integrals in `[0, ∞]`,
so "finite or infinite" is literal. `x⁺ = max x 0`. -/
theorem lemma_9_4 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(κ : ℝ) (hκ : 0 < κ) (X Y : Ω → ℝ) (hX : Measurable X) (hY : Measurable Y) :
(∫⁻ t in Set.Ioi (0 : ℝ),
ENNReal.ofReal (|P.real {ω | t < X ω} - P.real {ω | t < Y ω}| * t ^ (κ - 1))
≤ ENNReal.ofReal κ⁻¹ *
∫⁻ ω, ENNReal.ofReal |(max (X ω) 0) ^ κ - (max (Y ω) 0) ^ κ| ∂P) ∧
(Integrable (fun ω => (max (X ω) 0) ^ κ - (max (Y ω) 0) ^ κ) P →
IntegrableOn (fun t : ℝ => (P.real {ω | t < X ω} - P.real {ω | t < Y ω}) * t ^ (κ - 1))
(Set.Ioi 0) ∧
∫ t in Set.Ioi (0 : ℝ), (P.real {ω | t < X ω} - P.real {ω | t < Y ω}) * t ^ (κ - 1)
= κ⁻¹ * ∫ ω, ((max (X ω) 0) ^ κ - (max (Y ω) 0) ^ κ) ∂P) := by sorry
end GoldieRenewal.Kesten
Source
Goldie, Implicit renewal theory and tails of solutions of random equations, Ann. Appl. Probab. 1(1):126–166 (1991), DOI 10.1214/aoap/1177005985, p. 149, Lemma 9.4, (9.17)–(9.18)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.