Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 2.5, (2.44), p. 767 — P(S_n ≥ x) ≤ Π(b_j(x/n) + b_gj)exp{−ng(x/n)} for g with positive nondecreasing g′

Proved
NagaevLD.GenMoment.theorem_2_5

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

generalized-momentslarge-deviationsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1sums-of-independent-random-variables

Let n≥1n\ge1n≥1 and let X1,…,XnX_1,\dots,X_nX1​,…,Xn​ be independent real random variables with distribution functions Fj(u)=P(Xj<u)F_j(u)=P(X_j<u)Fj​(u)=P(Xj​<u), and Sn=X1+⋯+XnS_n=X_1+\dots+X_nSn​=X1​+⋯+Xn​. Let g:R→Rg:\mathbb R\to\mathbb Rg:R→R have a positive nondecreasing derivative g′g'g′, and put

bgj=∫u≥0eg(u) dFj(u),bj(s)=eg(0)∫u<0eg′(s)u dFj(u).b_{gj}=\int_{u\ge 0}e^{g(u)}\,dF_j(u),\qquad b_j(s)=e^{g(0)}\int_{u<0}e^{g'(s)u}\,dF_j(u).bgj​=∫u≥0​eg(u)dFj​(u),bj​(s)=eg(0)∫u<0​eg′(s)udFj​(u).

Assume each bgjb_{gj}bgj​ is finite. Then for any x>0x>0x>0,

P(Sn≥x) ≤ ∏j=1n(bj(x/n)+bgj)exp⁡{−ng(x/n)}.P(S_n\ge x)\ \le\ \prod_{j=1}^n\big(b_j(x/n)+b_{gj}\big)\exp\{-ng(x/n)\}.P(Sn​≥x) ≤ j=1∏n​(bj​(x/n)+bgj​)exp{−ng(x/n)}.

This bound, obtained by S. K. Sakoyan and Nagaev, controls the upper tail of a sum of independent summands through the generalized moments bgjb_{gj}bgj​, which can be much smaller than power moments when the right tails decay like e−ge^{-g}e−g. For fixed nnn and x→∞x\to\inftyx→∞, ∏j(bj(x/n)+bgj)→∏jbgj\prod_j(b_j(x/n)+b_{gj})\to\prod_j b_{gj}∏j​(bj​(x/n)+bgj​)→∏j​bgj​, so the bound has the exponential order e−ng(x/n)e^{-ng(x/n)}e−ng(x/n).

Formalization Note The summands are measurable and mutually independent (iIndepFun), indexed by Fin n. The finiteness of each bgjb_{gj}bgj​ is an explicit integrability hypothesis: the page treats bgjb_{gj}bgj​ as a finite factor, and Lean's integral of a non-integrable function is 000, which would make the bound false rather than trivial. The paper defines bgib_{gi}bgi​ over u>0u>0u>0 (p. 759); the formalization integrates over u≥0u\ge 0u≥0, the reading the proof uses, because with u>0u>0u>0 the inequality fails (see the definition file). The hypotheses on ggg — differentiability, g′>0g'>0g′>0, g′g'g′ nondecreasing — are imposed on [0,∞)[0,\infty)[0,∞) only, the region where ggg and g′g'g′ are evaluated; they imply that ggg is nondecreasing with g(u)→∞g(u)\to\inftyg(u)→∞, the standing assumption of §2. The paper's implicit n≥1n\ge1n≥1 is explicit in Lean, avoiding x/0x/0x/0.

Preamble
import Mathlib
import Definitions.Def_NagaevLD_GenMoment_Setting

open MeasureTheory ProbabilityTheory
Formal statement
namespace NagaevLD.GenMoment

/-- Theorem 2.5, (2.44), p. 767 (Nagaev–Sakoyan): if `g` has a positive nondecreasing
derivative `g′`, then for any `x > 0`,
`P(S_n ≥ x) ≤ ∏_{j=1}^n (b_j(x/n) + b_gj) · exp{−n g(x/n)}`. -/
theorem theorem_2_5 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
    (n : ℕ) (hn : 0 < n) (X : Fin n → Ω → ℝ) (hXm : ∀ j, Measurable (X j)) (hind : iIndepFun X P)
    (g g' : ℝ → ℝ) (hg : ∀ u : ℝ, 0 ≤ u → HasDerivAt g (g' u) u)
    (hg'pos : ∀ u : ℝ, 0 ≤ u → 0 < g' u) (hg'mono : MonotoneOn g' (Set.Ici 0))
    (hbg : ∀ j, IntegrableOn (fun ω => Real.exp (g (X j ω))) {ω | 0 ≤ X j ω} P)
    (x : ℝ) (hx : 0 < x) :
    P.real {ω | x ≤ NagaevLD.FukNagaev.S n X ω}
      ≤ (∏ j, (bj P X g g' j (x / n) + bg P X g j)) * Real.exp (-(n * g (x / n))) := by sorry

end NagaevLD.GenMoment
Source
Nagaev, Large deviations of sums of independent random variables, Ann. Probab. 7 (1979), p. 767, Theorem 2.5, (2.44); b_gi from p. 759
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me