§3, p. 14 — with v_0 = 0 and v_i0 = 0, offering only the highest-revenue product over all nests is optimal
ProvedNestedLogitVariants.Competitive.optimal_of_v0_eq_zeroConsider the nested logit assortment problem (2) in which no customer leaves without considering a nest, , and every nest is fully captured, for all nests . Products in each nest are ordered by revenue, , and . Let be a nest whose top product has the largest revenue over all nests, . Then the assortment that offers only product in nest and nothing in any other nest,
is an optimal solution to problem (2).
This is the case that §3 sets aside with its "without loss of generality ": in it, the conclusion of Theorem 4 holds directly, since the optimal assortment above is nested by revenue in every nest.
Formalization Note Products are indexed , so the paper's product is index and is nbr n 1. The hypothesis of §3 is not needed and is omitted, which makes the statement stronger. The paper's further remark that this product "would be purchased with probability one" is not stated. Standing assumptions and pins as in the model definition.
import Mathlib import Definitions.Def_NestedLogitVariants_Competitive_Model
namespace NestedLogitVariants.Competitive
/-- §3, p. 14 (the case `v_0 = 0` set aside by the section's "without loss of generality"):
when `v_0 = 0` and `v_{i0} = 0` for every nest, offering only the product with the largest revenue
`max_{l ∈ M} r_{l1}` (here product `0` of nest `i`) and nothing in the other nests is optimal. -/
theorem optimal_of_v0_eq_zero {ι : Type*} [Fintype ι] [DecidableEq ι] {n : ℕ} (I : Instance ι n) (hI : I.Standing)
(hfc : ∀ i, I.vnp i = 0) (hv0 : I.v0 = 0) (hn : 0 < n) (i : ι)
(hi : ∀ l, I.r l ⟨0, hn⟩ ≤ I.r i ⟨0, hn⟩) :
IsOptimal I (fun l => if l = i then nbr n 1 else ∅) := by sorry
end NestedLogitVariants.Competitive
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.