CVaR upper-bounds VaR for a finite-support law
ProvedDataDrivenRO.Discrete.var_le_cvarconditional-value-at-riskp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1risk-measuresvalue-at-risk
Let assign masses to the listed support vectors . For every and direction , conditional value at risk at tail level bounds value at risk at the same level:
The same conditional-value-at-risk function is convex as a function of the direction . The inequality is used in the first line of the proof of Theorem 4.
Formalization Note is the finitely supported point-mass law. The simplex hypothesis makes it a probability law and prevents the real infima from taking default values.
Preamble
import Mathlib import Definitions.Def_DataDrivenRO_Discrete_Setting
Formal statement
namespace DataDrivenRO.Discrete
/-- Conditional Value at Risk is an upper bound on Value at Risk for the same finite-support
law and tail level. Sentence following (11), p. 12. -/
theorem var_le_cvar {d n : ℕ} (a : Fin n → (Fin d → ℝ)) (p : Fin n → ℝ)
(hp : p ∈ stdSimplex ℝ (Fin n)) (ε : ℝ) (hε0 : 0 < ε) (hε1 : ε < 1)
(v : Fin d → ℝ) :
VaR a p ε v ≤ CVaR a p ε v ∧
ConvexOn ℝ Set.univ (fun w : Fin d → ℝ => CVaR a p ε w) := by sorry
end DataDrivenRO.Discrete
Source
Bertsimas, Gupta & Kallus, Data-Driven Robust Optimization, arXiv:1401.0212v2, sentence following (11), p. 12; proof of Theorem 4, p. ec2
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.