Corollary 1.8 (1.24), p. 753 — EX_i = 0, A_t⁺ < ∞, t ≥ 2: P(S_n ≥ x) ≤ (1 + 2/t)^tA_t⁺x^{−t} + exp{−2(t + 2)^{−2}e^{−t}x²/B_n²}
ProvedNagaevLD.FukNagaev.corollary_1_8Let be independent random variables with and finite variances , , and suppose for some . Then for every ,
This is the form in which the Fuk–Nagaev inequality is usually quoted: the tail of a sum of centred independent variables is at most a polynomial term, governed by the -th moments of the positive parts, plus a Gaussian term governed by the variance. The page obtains it from (1.23) with and Chebyshev's inequality.
Formalization Note is the sum of Mathlib's variance (X i) P, which is for a variable outside ; the finiteness of every variance is therefore assumed (). On the page an infinite makes the exponential term and the bound trivial. If , the exponential term is written as the page's limiting value (); Lean's would otherwise turn it into . is integrability of on . Powers with exponent are real powers.
import Mathlib import Definitions.Def_NagaevLD_FukNagaev_Setting
namespace NagaevLD.FukNagaev
open MeasureTheory ProbabilityTheory
theorem corollary_1_8 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
{n : ℕ} (X : Fin n → Ω → ℝ) (hmeas : ∀ i, Measurable (X i)) (hindep : iIndepFun X P)
(x : ℝ) (hx : 0 < x) (t : ℝ) (ht : 2 ≤ t)
(hL2 : ∀ i, MemLp (X i) 2 P) (hmean : ∀ i, ∫ ω, X i ω ∂P = 0)
(hAt : ∀ i, IntegrableOn (fun ω => |X i ω| ^ t) {ω | 0 ≤ X i ω} P) :
P.real {ω | x ≤ S n X ω} ≤
(1 + 2 / t) ^ t * Atplus P X t * x ^ (-t) +
(if Bn2 P X = 0 then 0
else Real.exp (-(2 * (t + 2) ^ (-2 : ℝ) * Real.exp (-t) * x ^ 2 / Bn2 P X))) := by sorry
end NagaevLD.FukNagaev
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.