Endpoint comparison of Gini means on a bounded interval
Abstract
A three-point rational witness refutes endpoint characterization of Gini mean comparisons.
Grünwald and Páles, Comparison of Gini means with fixed number of variables, arXiv:2408.07658v1, §4, introductory paragraph, verbatim: “We conclude this paper by formulating two conjectures that we have not been able to verify. The first conjecture seems to be a natural extension of the formula for Γ2(I), if 0 < inf I < sup I < ∞.” Conjecture 1 states verbatim: “Conjecture 1. Let n ≥ 3 and let I be a subinterval of R+ such that 0 < a := inf I < sup I =: b < ∞. Then Γn(I) = {((r, s), (p, q)) ∈ R² × R² | r + s ≤ p + q, G[n]r,s(a, …, a, b) ≤ G[n]p,q(a, …, a, b), and G[n]r,s(a, b, …, b) ≤ G[n]p,q(a, b, …, b)}.”
Definition 1.1 (The distinct-parameter Gini mean).
Formalization. D5/S3/ArithSums/GiniMeanEndpointComparison.giniMean (✓ std3).
Citation. R. Grünwald; Zs. Páles (2024). Comparison of Gini means with fixed number of variables. URL: https://arxiv.org/abs/2408.07658v1.
Commentary.
For x : Fin n → ℝ, giniMean r s x is the displayed real-power expression. For positive coordinates and r≠s it is the source’s Gini mean. The Lean definition is a total real-valued expression; the equal-parameter mean is not encoded.
Definition 1.2 (One upper endpoint).
Formalization. D5/S3/ArithSums/GiniMeanEndpointComparison.lowTuple (✓ std3).
Source. Repository-derived.
Acknowledgement. R. Grünwald; Zs. Páles (2024). Comparison of Gini means with fixed number of variables. URL: https://arxiv.org/abs/2408.07658v1.
Commentary.
The index i ranges over Fin n. The value val i is its natural-number index. The tuple places b at the last index and a elsewhere, so for n=3 it is (a,a,b).
Definition 1.3 (One lower endpoint).
Formalization. D5/S3/ArithSums/GiniMeanEndpointComparison.highTuple (✓ std3).
Source. Repository-derived.
Acknowledgement. R. Grünwald; Zs. Páles (2024). Comparison of Gini means with fixed number of variables. URL: https://arxiv.org/abs/2408.07658v1.
Commentary.
The tuple places a at index zero and b elsewhere, so for n=3 it is (a,b,b). Both tuple definitions also apply to n=0, when there are no indices.
Definition 1.4 (Sufficiency of the two extreme endpoint tests).
Formalization. D5/S3/ArithSums/GiniMeanEndpointComparison.claim (✓ std3).
Source. Repository-derived.
Acknowledgement. R. Grünwald; Zs. Páles (2024). Comparison of Gini means with fixed number of variables. URL: https://arxiv.org/abs/2408.07658v1.
Commentary.
The claim is the proposed ⊇ inclusion of Conjecture 1 for closed intervals [a,b] and distinct parameters r≠s and p≠q. It quantifies over n≥3 and 0<a<b. The parameter-sum inequality and both endpoint comparisons are required to imply comparison at every tuple whose coordinates lie in [a,b]. This restricted implication is enough to refute the full conjectured equality.
Theorem 1.5 (The extreme endpoint tests are insufficient).
Proof. Machine-checked in Lean as D5/S3/ArithSums/GiniMeanEndpointComparison.result (✓ std3). ∎
Resolves. Problems/grunwald-pales-2024-gini-mean-comparison (refuted) by D5/S3/ArithSums/GiniMeanEndpointComparison.result.
Source. Repository-derived.
Acknowledgement. R. Grünwald; Zs. Páles (2024). Comparison of Gini means with fixed number of variables. URL: https://arxiv.org/abs/2408.07658v1.
Commentary.
Take n=3, a=1, b=3, (r,s)=(3,−1), and (p,q)=(0,2). Both parameter sums equal 2. The lemmas gini_low_rs and gini_low_pq give fourth powers 87/7 and 121/9 at (1,1,3). The lemmas gini_high_rs and gini_high_pq give 33 and 361/9 at (1,3,3). The nonnegativity lemmas and le_fourth establish both endpoint comparisons. The private tuple x=(1,1,5/3) has all coordinates in [1,3]. The lemmas gini_x_rs and gini_x_pq give fourth powers 895/351 and 1849/729 there. Their order reverses: 895·729=652455>648999=1849·351. Applying the claim to this tuple yields a false fourth-power inequality. Thus the proposed ⊇ inclusion, and hence Conjecture 1, fails. As mathematical reasoning, homogeneity scales x to (a,a,5a/3), an endpoint configuration at a smaller ratio. For these parameters the comparison fails at ratio 5/3 but holds at ratio 3, so the extreme ratio does not control the intermediate ratios. A candidate correction requires both endpoint tests for every t∈(1,b/a]; its sufficiency is not proved. Further integer-parameter violations, including (−2,0,4,−6) on [1,4] at (1,13/8,13/8), are computed numerical observations and are not formalized. Conjecture 2 for inf I=0 or sup I=∞, and the true bounded-interval description of Γn(I), are not settled by this result. The journal version’s text is ASSUMED-UNVERIFIED; the quotations refer to arXiv:2408.07658v1, §4.
References
- Truth anchor:
D5/S3/ArithSums/GiniMeanEndpointComparison.claim - Truth anchor:
D5/S3/ArithSums/GiniMeanEndpointComparison.giniMean - Truth anchor:
D5/S3/ArithSums/GiniMeanEndpointComparison.highTuple - Truth anchor:
D5/S3/ArithSums/GiniMeanEndpointComparison.lowTuple - Truth anchor:
D5/S3/ArithSums/GiniMeanEndpointComparison.result