Proof of Theorem 6, p. 801 — integrating over
ProvedBalkemaDeHaan.ParetoBounds.integration_displayLet be a distribution function on which has a positive density for , and let satisfy
Then for every and every ,
This is the integrated form of the hazard-rate bounds: the middle integral is the increase of the cumulative hazard between and , and the outer integrals equal . Applying the monotone map turns it into Theorem 6.
Formalization Note is cdf μ for a probability measure , and the density is an explicit function with differentiable at every with derivative , the derivative being taken within (so only a right derivative is required at itself; for this is the ordinary derivative). The integrals are oriented interval integrals; the hypotheses force (at the ratio is ), so . The integrand is not assumed integrable: under the hypotheses it is bounded on and measurable there, so the statement does not rest on Lean's value for non-integrable functions (that value would make the first inequality false, not trivially true).
import Mathlib open MeasureTheory ProbabilityTheory
namespace BalkemaDeHaan.ParetoBounds
/-- Proof of Theorem 6, p. 801 (Balkema–de Haan 1974): integrating the hazard bounds
`α₁ ≤ t F'(t)/(1 - F(t)) ≤ α₂` between `t` and `(1 + x) t`, `x > 0`, gives
`α₁ ∫ du/u ≤ ∫ F'(u)/(1 - F(u)) du ≤ α₂ ∫ du/u`. The density `F' = f` is a (right,
at `t₀`) derivative of `F = cdf μ` on `[t₀, ∞)`. -/
theorem integration_display (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 : ℝ) (ht : t₀ ≤ t) (x : ℝ) (hx : 0 < x) :
α₁ * ∫ u in t..(1 + x) * t, 1 / u ≤ ∫ u in t..(1 + x) * t, f u / (1 - cdf μ u) ∧
∫ u in t..(1 + x) * t, f u / (1 - cdf μ u) ≤ α₂ * ∫ u in t..(1 + x) * t, 1 / u := by sorry
end BalkemaDeHaan.ParetoBounds
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.