Theorem 4.1 (part 1) — R =_L Q + MR has a unique stationary law
OpenGoldieRenewal.Kesten.existence_uniquenessLet and be random variables with joint law on (no independence and no sign condition is assumed). Suppose satisfies the conditions of Lemma 2.2 for some : , , and the conditional law of given is nonarithmetic. Suppose also that (4.2) .
Then there is a unique probability law on for satisfying the random difference equation
That is, is the law of when , and every probability law with this property equals .
This is the first assertion of Theorem 4.1 (Kesten's theorem in dimension one). It is the classical existence and uniqueness of the stationary law of the affine recursion . Under the hypotheses, Lemma 2.2 gives , and gives . The stationary law is then the law of the a.s. convergent perpetuity .
Formalization Note A law for satisfying (1.1) is a probability measure with rdeOperator μ ρ = ρ, i.e. the image of under equals . Uniqueness is among all probability laws on .
import Mathlib import Definitions.Def_GoldieRenewal_Implicit_CramerConditions import Definitions.Def_GoldieRenewal_Kesten_RandomDifferenceEquation
namespace GoldieRenewal.Kesten
open MeasureTheory ProbabilityTheory
open scoped ENNReal
/-- **Theorem 4.1, part 1: existence and uniqueness of the law of `R`** (Goldie, *Implicit renewal
theory and tails of solutions of random equations*, Ann. Appl. Probab. 1(1) (1991), p. 135,
Theorem 4.1: "there is a unique law for `R` satisfying (1.1)"). Let `(Q, M)` have joint law `μ`,
let `M` satisfy the conditions of Lemma 2.2 for some `κ > 0`, and suppose (4.2) `E|Q|^κ < ∞`.
Then there is exactly one probability law `ρ` on `ℝ` with `R =_L Q + MR`, `R` independent of
`(Q, M)`.
**Formalization Note** A law for `R` satisfying (1.1) is a probability measure `ρ` with
`rdeOperator μ ρ = ρ`; uniqueness is among all probability laws on `ℝ`. This is the first
conjunct of `GoldieRenewal.Kesten.theorem_4_1`. -/
theorem existence_uniqueness (κ : ℝ) (μ : Measure (ℝ × ℝ)) [IsProbabilityMeasure μ]
(hM : GoldieRenewal.Implicit.CramerConditions κ (μ.map Prod.snd))
(hQ : ∫⁻ p, ENNReal.ofReal (|p.1| ^ κ) ∂μ < ∞) :
∃ ρ : ProbabilityMeasure ℝ,
rdeOperator μ (ρ : Measure ℝ) = ρ ∧
∀ ρ' : ProbabilityMeasure ℝ, rdeOperator μ (ρ' : Measure ℝ) = ρ' → ρ' = ρ := by sorry
end GoldieRenewal.Kesten