App. A, p. ec1 — W₋₁(x) is the solution of we^w = x in (−∞, −1]; W₋₁(−1/e) = −1, W₋₁(x) → −∞ as x → 0⁻, W₋₁ non-increasing (corrected: one solution at −1/e)
ProvedValuePersPricing.Deviation.lambertWm1_factslambert-wp2o-batch-pfp2ap2o-correctedp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1
Let denote the negative real branch of the Lambert function. Then:
- for every the equation has exactly two real solutions, and for its only real solution is ;
- for every , , , and is the only solution of in ;
- ;
- is non-increasing on .
These are the properties of used throughout the proofs of Theorem 2 and Lemmas 2, EC.1 and EC.2: the defining equation turns the integral inequalities into closed-form bounds, and monotonicity lets one apply to both sides of an inequality.
Formalization Note lambertWm1 is defined by choice with the junk value off ; the statement never evaluates it there. The limit is one-sided, from the left, because is defined only on .
Differences from the source:
- correction: "When x ∈ [−1/e, 0), this equation has two distinct real solutions" → two solutions for x ∈ (−1/e, 0) and the single solution −1 at x = −1/e; reason: at x = −1/e, w e^w attains its minimum −1/e only at w = −1, so the printed claim is false there.
- encoding: lim_{x→0} W₋₁(x) = −∞ → limit along x → 0 from the left; reason: W₋₁ is defined only on [−1/e, 0).
- generalization: App. A, p. ec1 states the equation, the range, the value at −1/e and the limit → Lean also states that W₋₁ is non-increasing on [−1/e, 0); reason: the paper asserts this on pp. ec2, ec4 and ec6 and uses it in every Lambert-W step; the extra conjunct only adds to the App. A statement.
Preamble
import Mathlib import Definitions.Def_ValuePersPricing_Deviation_Model open MeasureTheory Set Filter Topology
Formal statement
namespace ValuePersPricing.Deviation
/-- App. A (p. ec1): `w e^w = x` has exactly two real solutions for `x ∈ (−1/e, 0)` and the single
solution `−1` at `x = −1/e` (corrected: the printed "two distinct real solutions" for
`x ∈ [−1/e, 0)` fails at `−1/e`); on `[−1/e, 0)`, `W₋₁(x)` is the unique solution in `(−∞, −1]`;
`W₋₁(−1/e) = −1`; `W₋₁(x) → −∞` as `x → 0⁻`; and `W₋₁` is non-increasing on `[−1/e, 0)`
(asserted on pp. ec2, ec4, ec6). -/
theorem lambertWm1_facts :
(∀ x ∈ Ioo (-1 / Real.exp 1) 0, {w : ℝ | w * Real.exp w = x}.ncard = 2) ∧
{w : ℝ | w * Real.exp w = -1 / Real.exp 1} = {-1} ∧
(∀ x : ℝ, -1 / Real.exp 1 ≤ x → x < 0 →
lambertWm1 x ≤ -1 ∧ lambertWm1 x * Real.exp (lambertWm1 x) = x ∧
∀ w : ℝ, w ≤ -1 → w * Real.exp w = x → w = lambertWm1 x) ∧
lambertWm1 (-1 / Real.exp 1) = -1 ∧
Tendsto lambertWm1 (𝓝[<] 0) atBot ∧
AntitoneOn lambertWm1 (Ico (-1 / Real.exp 1) 0) := by sorry
end ValuePersPricing.Deviation
Source
Elmachtoub, Gupta & Hamilton, The Value of Personalized Pricing, SSRN 3127719 (review version, PDF sha256 635e22eb…), App. A, p. ec1 (PDF 33); monotonicity: pp. ec2, ec4, ec6 (PDF 34, 36, 38)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.