Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 6 — α1≤tF′(t)/(1−F(t))≤α2\alpha_1 \le tF'(t)/(1-F(t)) \le \alpha_2α1​≤tF′(t)/(1−F(t))≤α2​ for t≥t0t \ge t_0t≥t0​ gives Γα1(x)≤P{(X−t)/t≤x∣X>t}≤Γα2(x)\Gamma_{\alpha_1}(x) \le P\{(X-t)/t \le x \mid X>t\} \le \Gamma_{\alpha_2}(x)Γα1​​(x)≤P{(X−t)/t≤x∣X>t}≤Γα2​​(x)

Proved
BalkemaDeHaan.ParetoBounds.theorem_6

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

extreme-value-theoryhazard-ratelimit-theoremp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1residual-life

Let FFF be a distribution function on R\mathbb RR which has a positive density F′F'F′ for t≥t0t \ge t_0t≥t0​, and let α1\alpha_1α1​ and α2\alpha_2α2​ be positive real numbers such that

α1≤tF′(t)1−F(t)≤α2for t≥t0.\alpha_1 \le \frac{t F'(t)}{1 - F(t)} \le \alpha_2 \qquad \text{for } t \ge t_0 .α1​≤1−F(t)tF′(t)​≤α2​for t≥t0​.

If XXX is a random variable with distribution FFF, then for every t≥t0t \ge t_0t≥t0​ and every real xxx,

Γα1(x)  ≤  P{X−tt≤x  ∣  X>t}  ≤  Γα2(x),\Gamma_{\alpha_1}(x) \;\le\; P\Big\{\frac{X - t}{t} \le x \;\Big|\; X > t\Big\} \;\le\; \Gamma_{\alpha_2}(x),Γα1​​(x)≤P{tX−t​≤x​X>t}≤Γα2​​(x),

where Γα(x)=1−(1+x)−α\Gamma_\alpha(x) = 1 - (1+x)^{-\alpha}Γα​(x)=1−(1+x)−α for x≥0x \ge 0x≥0 and Γα(x)=0\Gamma_\alpha(x) = 0Γα​(x)=0 for x<0x < 0x<0.

The quantity tF′(t)/(1−F(t))tF'(t)/(1 - F(t))tF′(t)/(1−F(t)) is the age times the hazard rate. When it equals a constant α\alphaα the tail is exactly Pareto, 1−F(t)=Ct−α1 - F(t) = C t^{-\alpha}1−F(t)=Ct−α, and both bounds are equalities. The theorem is a finite-ttt, two-sided version of von Mises' condition tF′(t)/(1−F(t))→αtF'(t)/(1-F(t)) \to \alphatF′(t)/(1−F(t))→α: bounds on the scaled hazard rate give explicit bounds on the residual life distribution, uniformly for all ages t≥t0t \ge t_0t≥t0​. The smaller constant α1\alpha_1α1​ (heavier tail) gives the lower bound.

Formalization Note FFF is cdf μ for a probability measure μ\muμ (the law of XXX), and the probability is Ft(xt)=μ((t,t+xt])/μ((t,∞))F_t(xt) = \mu((t, t + xt])/\mu((t, \infty))Ft​(xt)=μ((t,t+xt])/μ((t,∞)). "Positive density F′F'F′ for t≥t0t \ge t_0t≥t0​" is an explicit function fff with FFF differentiable at every t≥t0t \ge t_0t≥t0​ with derivative f(t)>0f(t) > 0f(t)>0, the derivative being taken within [t0,∞)[t_0, \infty)[t0​,∞) (a right derivative at t0t_0t0​, the ordinary derivative for t>t0t > t_0t>t0​); this is the weakest reading of the page and admits the Pareto law with t0=1t_0 = 1t0​=1, whose FFF has no two-sided derivative at 111. No hypothesis is added: t0>0t_0 > 0t0​>0 and F(t)<1F(t) < 1F(t)<1 for all ttt are consequences of the stated hypotheses (at t≤0t \le 0t≤0 the ratio is not ≥α1>0\ge \alpha_1 > 0≥α1​>0; a positive derivative on [t0,∞)[t_0,\infty)[t0​,∞) makes FFF strictly increasing there), so Lean's conventions t⋅f(t)/0=0t\cdot f(t)/0 = 0t⋅f(t)/0=0 and a/0=0a/0 = 0a/0=0 never act at the ages concerned. The conclusion holds for all real xxx, as printed; for x≤0x \le 0x≤0 all three quantities are 000.

Preamble
import Mathlib
import Definitions.Def_BalkemaDeHaan_ParetoBounds_ResidualLife
import Definitions.Def_BalkemaDeHaan_ParetoBounds_GammaLaw

open MeasureTheory ProbabilityTheory
Formal statement
namespace BalkemaDeHaan.ParetoBounds

/-- Theorem 6, p. 801 (Balkema–de Haan 1974): if `F = cdf μ` has a positive density `f`
on `[t₀, ∞)` with `α₁ ≤ t f(t)/(1 - F(t)) ≤ α₂` for `t ≥ t₀`, then for every `t ≥ t₀`
and every real `x`,
`Γ_{α₁}(x) ≤ P{(X - t)/t ≤ x | X > t} ≤ Γ_{α₂}(x)`, where
`P{(X - t)/t ≤ x | X > t} = F_t(x t)`. -/
theorem theorem_6 (μ : Measure ℝ) [IsProbabilityMeasure μ] (f : ℝ → ℝ)
    (t₀ α₁ α₂ : ℝ) (hα₁ : 0 < α₁) (hα₂ : 0 < α₂)
    (hdens : ∀ t ≥ t₀, HasDerivWithinAt (cdf μ) (f t) (Set.Ici t₀) t ∧ 0 < f t)
    (hbounds : ∀ t ≥ t₀, α₁ ≤ t * f t / (1 - cdf μ t) ∧ t * f t / (1 - cdf μ t) ≤ α₂) :
    ∀ t ≥ t₀, ∀ x : ℝ,
      GammaLaw α₁ x ≤ residualLife μ t (x * t) ∧ residualLife μ t (x * t) ≤ GammaLaw α₂ x := by sorry

end BalkemaDeHaan.ParetoBounds
Source
Balkema, de Haan, Residual Life Time at Great Age, Ann. Probab. 2 (1974), p. 801 (PDF p. 10), §4, Theorem 6
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