Lemma 3, p. 14 — removing a product with r_ij < γ_i Z + (1 − γ_i) R_i(S_i) strictly increases revenue
ProvedNestedLogitVariants.Competitive.lemma_3Assume and for every nest , and . Let be any assortment, with expected revenue . Suppose that in some nest the conditional revenue satisfies , and that some product has a small revenue:
Then removing product from yields an assortment whose expected revenue is strictly larger than .
The lemma provides the mechanism by which products with small revenues are removed from an assortment without loss; combined with Proposition 2 it gives the revenue threshold satisfied by every product of an optimal assortment.
Formalization Note The standing assumptions of §3 (p. 13), and for every nest, are hypotheses, together with those of §1 and the disclosed pins of the model definition (, , , revenues ordered within each nest). The positivity is the section's own "without loss of generality" assumption (p. 14), stated as a hypothesis here because the argument uses it.
import Mathlib import Definitions.Def_NestedLogitVariants_Competitive_Model
namespace NestedLogitVariants.Competitive
/-- Lemma 3, p. 14: if `Z = Π(S_1, …, S_m)` and a product `j ∈ S_i` satisfies
`r_ij < γ_i Z + (1 − γ_i) R_i(S_i)` and `R_i(S_i) ≥ Z`, then removing `j` from `S_i` yields a
strictly larger expected revenue than `Z`. -/
theorem lemma_3 {ι : Type*} [Fintype ι] [DecidableEq ι] {n : ℕ} (I : Instance ι n) (hI : I.Standing)
(hγ : ∀ i, I.γ i ≤ 1) (hfc : ∀ i, I.vnp i = 0) (hv0 : 0 < I.v0)
(S : ι → Finset (Fin n)) (i : ι) (j : Fin n) (hj : j ∈ S i)
(hr : I.r i j < I.γ i * revenue I S + (1 - I.γ i) * R I i (S i))
(hR : revenue I S ≤ R I i (S i)) :
revenue I S < revenue I (Function.update S i ((S i).erase j)) := by sorry
end NestedLogitVariants.Competitive
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.