Lemma 4 — Axiom 6 iff the quadratic-program minimum is zero
ProvedMcFadden1974.QPTest.axiom6_iff_qp_min_zeroconditional-logitlinear-optimizationmaximum-likelihoodp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Consider a finite conditional-logit experiment satisfying Axiom 5. Let be the set of vectors with every . Then Axiom 6 holds exactly when the quadratic program attains zero:
This converts the existence condition for the conditional-logit maximum likelihood estimate into a finite quadratic-programming test.
Formalization Note The minimum is required to be attained, as in the paper; a mere infimum of zero would state less. The data convention includes at least one observed choice in each trial.
Preamble
import Definitions.Def_McFadden1974_QPTest_ChoiceData set_option autoImplicit false
Formal statement
namespace McFadden1974.QPTest
/-- McFadden (1974), p. 117 (PDF 13), Lemma 4 and equation (22).
Under Axiom 5, Axiom 6 holds exactly when the quadratic program attains a
minimum of zero. `IsLeast` retains attainment, unlike an infimum equation.
The finite data model includes at least two alternatives and at least one
observed choice per trial; Axiom 5 uses the equivalent difference span form. -/
theorem axiom6_iff_qp_min_zero {N K : ℕ} (d : ChoiceData N K)
(h5 : d.Axiom5) :
d.Axiom6 ↔
IsLeast ((fun y : EuclideanSpace ℝ (Fin K) => ‖y‖ ^ 2) '' d.qpFeasible) 0 := by sorry
end McFadden1974.QPTest
Source
McFadden, Conditional Logit Analysis of Qualitative Choice Behavior, in P. Zarembka (ed.), Frontiers in Econometrics, Academic Press (1974), p. 117, Lemma 4 and Equation (22) (PDF p. 13)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.