Corollary 2.4 — implicit renewal theorem with moment conditions (2.16)–(2.17) and constants (2.18)–(2.20)
ProvedGoldieRenewal.Kesten.corollary_2_4Let be a probability space, jointly measurable, and , real random variables. Suppose satisfies the random equation (2.1), , that satisfies the conditions of Lemma 2.2 for some — , , given nonarithmetic — and that is independent of . Let .
- Case 1, a.s. If (2.16), then
if (2.17), then (2.19). 2. Case 2, . If both (2.16) and (2.17) hold, then both limits hold with
This is the form of the implicit renewal theorem used for concrete random equations: the integrability conditions of Theorem 2.3 on tail differences are replaced by moment conditions on and , which can be checked directly.
Formalization Note "" is stated as , the reading the paper prescribes on p. 130 (it includes ). Independence of from is independence of from , with the product σ-algebra on functions . In (2.16) the printed connective is read as a minus sign.
import Mathlib import Definitions.Def_GoldieRenewal_Implicit_CramerConditions
namespace GoldieRenewal.Kesten
open MeasureTheory ProbabilityTheory Filter Topology
/-- **Corollary 2.4** (Goldie, *Implicit renewal theory and tails of solutions of random
equations*, Ann. Appl. Probab. 1(1) (1991), p. 130). Let `Ψ : ℝ × Ω → ℝ` be jointly measurable,
let `R` satisfy (2.1) (`Ψ(R) =_L R`), and let `M` satisfy the conditions of Lemma 2.2 with
`R` independent of `(Ψ, M)`. Write `m = E|M|^κ log|M|`.
* Case 1 (`M ≥ 0` a.s.): if (2.16) `E|(Ψ(R)⁺)^κ − ((MR)⁺)^κ| < ∞` then
`t^κ P(R > t) → C₊ = (κm)⁻¹ E((Ψ(R)⁺)^κ − ((MR)⁺)^κ)` (2.18); if (2.17)
`E|(Ψ(R)⁻)^κ − ((MR)⁻)^κ| < ∞` then `t^κ P(R < −t) → C₋ = (κm)⁻¹ E((Ψ(R)⁻)^κ − ((MR)⁻)^κ)` (2.19).
* Case 2 (`P(M < 0) > 0`): if both (2.16) and (2.17) hold, both limits hold with
`C₊ = C₋ = (2κm)⁻¹ E(|Ψ(R)|^κ − |MR|^κ)` (2.20).
**Formalization Note** (2.16) is printed with a glyph resembling "∼" in place of "−"; read "−"
(see (2.17)). "`P(R > t) ~ C₊ t^{−κ}`" is read, as the paper instructs on p. 130 (which covers
`C₊ = 0`), as `t^κ P(R > t) → C₊`. "`R` independent of `(Ψ, M)`" is `IndepFun` of `R` and
`ω ↦ (M ω, (t ↦ Ψ(t, ω)))`, with the product σ-algebra on `ℝ → ℝ`. Conditions (2.16)/(2.17) are
`Integrable` hypotheses, so the expectations in (2.18)–(2.20) are genuine. -/
theorem corollary_2_4 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(κ : ℝ) (Ψ : ℝ → Ω → ℝ) (hΨ : Measurable (fun p : ℝ × Ω => Ψ p.1 p.2))
(M R : Ω → ℝ) (hM : Measurable M) (hR : Measurable R)
(hCramer : GoldieRenewal.Implicit.CramerConditions κ (P.map M))
(hindep : IndepFun R (fun ω => (M ω, fun t => Ψ t ω)) P)
(hfix : P.map (fun ω => Ψ (R ω) ω) = P.map R) :
(P {ω | M ω < 0} = 0 →
(Integrable (fun ω => (max (Ψ (R ω) ω) 0) ^ κ - (max (M ω * R ω) 0) ^ κ) P →
Tendsto (fun t : ℝ => t ^ κ * P.real {ω | t < R ω}) atTop
(𝓝 ((κ * GoldieRenewal.Implicit.cramerMean κ (P.map M))⁻¹ *
∫ ω, ((max (Ψ (R ω) ω) 0) ^ κ - (max (M ω * R ω) 0) ^ κ) ∂P))) ∧
(Integrable (fun ω => (max (-Ψ (R ω) ω) 0) ^ κ - (max (-(M ω * R ω)) 0) ^ κ) P →
Tendsto (fun t : ℝ => t ^ κ * P.real {ω | R ω < -t}) atTop
(𝓝 ((κ * GoldieRenewal.Implicit.cramerMean κ (P.map M))⁻¹ *
∫ ω, ((max (-Ψ (R ω) ω) 0) ^ κ - (max (-(M ω * R ω)) 0) ^ κ) ∂P)))) ∧
(0 < P {ω | M ω < 0} →
Integrable (fun ω => (max (Ψ (R ω) ω) 0) ^ κ - (max (M ω * R ω) 0) ^ κ) P →
Integrable (fun ω => (max (-Ψ (R ω) ω) 0) ^ κ - (max (-(M ω * R ω)) 0) ^ κ) P →
Tendsto (fun t : ℝ => t ^ κ * P.real {ω | t < R ω}) atTop
(𝓝 ((2 * κ * GoldieRenewal.Implicit.cramerMean κ (P.map M))⁻¹ *
∫ ω, (|Ψ (R ω) ω| ^ κ - |M ω * R ω| ^ κ) ∂P)) ∧
Tendsto (fun t : ℝ => t ^ κ * P.real {ω | R ω < -t}) atTop
(𝓝 ((2 * κ * GoldieRenewal.Implicit.cramerMean κ (P.map M))⁻¹ *
∫ ω, (|Ψ (R ω) ω| ^ κ - |M ω * R ω| ^ κ) ∂P))) := by sorry
end GoldieRenewal.Kesten
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.