Siegel Lemma 2.1: the averaged CDF at the mean is at least
Provedsiegel_cdf_average_ge_meanSiegel 2001, Lemma 2.1 (Median Bounds and their Application, J. Algorithms 38:184-236, p.4-5). Let be a nonnegative random variable with cumulative distribution function (so , the density, ) and finite mean , with total mass . Then the average value of the CDF over is at least :
The improper facts (total mass and mean ) are taken as hypotheses (standard probabilistic packaging). This is the reusable real-analysis core of Siegel's median machinery, used to prove the moustache-CDF value bound (Theorem 2.1) and thence the integer-mean binomial/Bernoulli median (Theorem 2.2 = Jogdeo-Samuels).
Preamble
import Mathlib.MeasureTheory.Integral.IntervalIntegral.IntegrationByParts import Mathlib.MeasureTheory.Integral.IntervalIntegral.FundThmCalculus import Mathlib.MeasureTheory.Integral.IntegralEqImproper import Mathlib.Analysis.SpecialFunctions.Integrals.Basic
Formal statement
theorem siegel_cdf_average_ge_mean (μ : ℝ) (hμ : 0 ≤ μ) (F f : ℝ → ℝ) (hF : ∀ x ∈ Set.Ici (0:ℝ), HasDerivAt F (f x) x) (hfnn : ∀ x ∈ Set.Ici (0:ℝ), 0 ≤ f x) (hF0 : F 0 = 0) (hf_int_loc : ∀ b : ℝ, IntervalIntegrable f MeasureTheory.volume 0 b) (hf_mass : MeasureTheory.IntegrableOn f (Set.Ioi (0:ℝ))) (htf_mass : MeasureTheory.IntegrableOn (fun t => t * f t) (Set.Ioi (0:ℝ))) (hmass : ∫ t in Set.Ioi (0:ℝ), f t = 1) (hmean : ∫ t in Set.Ioi (0:ℝ), t * f t = μ) : μ ≤ ∫ t in (0:ℝ)..(2*μ), F t := by sorry
Source
Siegel, A. (2001). Median Bounds and their Application. Journal of Algorithms 38(1):184-236, Lemma 2.1, p.4-5. Proof = single integration by parts of the CDF against (t-2mu).