Documentation

LeanPool.RareSubtreePatterns.RootDeclumping.LeafHeightMark

Leaf-height declumping: the root formula and the conditional mark bound #

Section 7.2 of the paper (source lines 3132–3181, 3236–3241; inventory U11 and Lemma lem:lh_conditional_mark). For the leaf-height marks G_k(T) = 1{λ_∅(T) ≥ k}, with κ = min{j ≥ 1 : μ_j > 0}, the declumped mark is Ghat_k(T) = 1{k_∅(T) ≥ κ + 1} 1{∃ u ∈ Γ_∅(T), λ_∅(T_u) ≥ k} (lhDeclumpedMark). Write π_k = P(λ_∅(𝒯) ≥ k) (piSeq, identified by bienayme_real_le_rootLeafHeight) and π̂_k = P(Ghat_k(𝒯) = 1).

The degree set B = {j : j ≥ κ + 1} of the leaf-height declumping (source line 3133).

Equations
Instances For

    The declumped leaf-height mark Ghat_k(T) = 1{k_∅(T) ≥ κ + 1} 1{∃ u ∈ Γ_∅(T), λ_∅(T_u) ≥ k} (source lines 3132–3137, with G_k(T) = 1{λ_∅(T) ≥ k} and B = {j ≥ κ + 1}).

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For

      c_B = c = 1 - κ μ_κ for B = {j ≥ κ + 1} (source line 3031).

      theorem PoissonDeclumping.RootDeclumping.lh_declumped_prob_eq {μ : PMF ℕ} (hc : IsCritical μ) (h1 : μ 1 ≠ 1) (k : ℕ) :

      U11, the leaf-height declumped root formula (source lines 3134–3141): π̂_k = ∑_{j ≥ κ + 1} μ_j (1 - (1 - π_k)^j).

      U11: π̂_k - c π_k = o(π_k) as k → ∞ (dominated convergence, source line 3142).

      U11: π̂_k / π_k → c (source line 3142).

      U11: if c = 1 - κ μ_κ > 0, then π̂_k ∼ c π_k (source line 3142).

      b = P(k_∅(𝒯) ≥ κ + 1) is the constant b_B of B = {j ≥ κ + 1} (source line 3236).

      If c > 0 then b = P(k_∅(𝒯) ≥ κ + 1) > 0 (source line 3236).

      π̂_k ≥ b π_k with b = P(k_∅(𝒯) ≥ κ + 1) (source lines 3236–3241).

      π̂_k ≤ c π_k ≤ π_k.

      Lemma lem:lh_conditional_mark #

      The event {Ghat_k(𝒯) = 1, λ_∅(𝒯) ≥ m + 1} in product form: root degree ≥ κ + 1, all children with leaf-height ≥ m, some child with leaf-height ≥ k.

      theorem PoissonDeclumping.RootDeclumping.pow_sub_pow_le_of_le {a b : ℝ} (hb0 : 0 ≤ b) (hba : b ≤ a) (ha1 : a ≤ 1) {d κ : ℕ} (hd : κ + 1 ≤ d) :
      a ^ d - b ^ d ≤ ↑d * (a - b) * a ^ κ

      a^d - b^d ≤ d (a - b) a^κ for 0 ≤ b ≤ a ≤ 1 and d ≥ κ + 1.

      Lemma lem:lh_conditional_mark (source lines 3148–3181): for a critical μ with μ_1 ≠ 1, there is one constant C > 0 such that, for all 0 ≤ m ≤ k, P(Ghat_k(𝒯) = 1, λ_∅(𝒯) ≥ m) ≤ C π_k π_m. One can take C = μ_κ⁻¹. (The paper states it in a subsection assuming c > 0; that assumption is not needed.)

      Lemma lem:lh_conditional_mark with π_k = P(λ_∅(𝒯) ≥ k) written as a probability.