Proof of Proposition 19, p. 876 — (Ṽ_2, Ṽ_3, …)/(1 − Ṽ_1) has the stick-breaking law with parameters (α, α + θ), independent of Ṽ_1
ProvedPoissonDirichlet.MaxDensity.stick_tailp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1poisson-dirichletstick-breaking
Let , , and let be independent, with stick-breaking sequence as in (4). Then the renormalised tail
has the same law as the stick-breaking sequence built from independent variables, that is, the law of ; and it is independent of .
This is the self-similarity of stick-breaking used in the proof of Proposition 19; the independence is what allows the conditioning on in (92).
Formalization Note. The page states only the identity of laws; the independence from is used in the next line of (92) and is part of Proposition 34, and is stated here as a second conjunct. The law of the stick-breaking sequence is expressed through measures of preimages under stick.
Preamble
import Mathlib import Definitions.Def_PoissonDirichlet_MaxDensity_Setting open MeasureTheory ProbabilityTheory Filter Topology
Formal statement
namespace PoissonDirichlet.MaxDensity
/-- Proof of Proposition 19, §5.1, p. 876 (Pitman–Yor 1997): "the consequence of (4) that the
`P_{α,θ}` distribution of `(Ṽ₂, Ṽ₃, …)/(1 - Ṽ₁)` is identical to the `P_{α,α+θ}`
distribution of `(Ṽ₁, Ṽ₂, …)`". In the setting of Definition 1 (0-based: `Ytil k` is
`Ỹ_{k+1}`), the sequence `k ↦ PoissonDirichlet.Ratio.stick Ỹ (k+1) / (1 - Ỹ₁)` has the law of `stick` under the
`(α, α + θ)` law of Definition 1, and it is independent of `Ṽ₁ = Ỹ₁` (the independence is
used in the second line of (92)). -/
theorem stick_tail {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(α θ : ℝ) (hα : 0 ≤ α) (hα1 : α < 1) (hθ : -α < θ)
(Ytil : ℕ → Ω → ℝ) (hYm : ∀ k, Measurable (Ytil k)) (hYind : iIndepFun Ytil P)
(hYlaw : ∀ k : ℕ, HasLaw (Ytil k) (betaMeasure (1 - α) (θ + ((k : ℝ) + 1) * α)) P) :
(∃ μ : Measure (ℕ → ℝ), PoissonDirichlet.Ratio.IsStickLaw α (α + θ) μ ∧
∀ s : Set (ℕ → ℝ), MeasurableSet s →
P ((fun ω k => PoissonDirichlet.Ratio.stick (fun i => Ytil i ω) (k + 1) / (1 - Ytil 0 ω)) ⁻¹' s) =
μ (PoissonDirichlet.Ratio.stick ⁻¹' s)) ∧
IndepFun (fun ω k => PoissonDirichlet.Ratio.stick (fun i => Ytil i ω) (k + 1) / (1 - Ytil 0 ω))
(Ytil 0) P := by sorry
end PoissonDirichlet.MaxDensity
Source
Pitman and Yor, The two-parameter Poisson–Dirichlet distribution derived from a stable subordinator, Ann. Probab. 25 (1997), p. 876, §5.1, proof of Proposition 19, sentence before (92)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.