REVIEW 3 major objections 4 minor 27 references
On the exact region between Chatterjee's rank correlation and Spearman's footrule
T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash
Pith's one-line read The attainable (ξ, ψ) region for stochastically increasing copulas is exactly x ≤ y ≤ √x, with the Fréchet family uniquely tracing the upper curve and ordinal sums of the independence copula tracing the diagonal.
desk verdict A clean, well-executed proof of the upper boundary ψ≤√ξ and the exact (ξ,ψ) region for stochastically increasing copulas; the main caveat is heavy reliance on one external theorem that is not re-derived. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The argument runs through three linked devices. First, the Markov-product identity $\xi(C)=\psi(C^{\top}\ast C)$, where $C^{\top}(u,v)=C(v,u)$ and $(C_1\ast C_2)(u,v)$ is the Markov product; this rewrites Chatterjee's coefficient as Spearman's footrule of a derived copula. Second, Lemma 1.1 characterizes a copula by its partial derivative $h_v(t)=\partial_1 C(t,v)$, subject to $\int_0^1 h_v(t)\,dt=v$ and monotonicity in $v$; in terms of $h$, $\psi$ is a linear functional while $\xi$ is a quadratic functional, turning the region problem into convex optimization over square-integrable functions. Third, for stochastically increasing copulas the diagonal inequality $(C^{\top}\ast C)(v,v)\le C(v,v)$ yields $\xi\le\psi$; the upper boundary then comes from a KKT analysis of the convex program, and the lower boundary of the SI region comes from this diagonal inequality together with continuity of $\xi$ along the convex segment between an ordinal-sum copula and a Fréchet copula.
What would settle it
Take any copula $C$ and compute, on a fine grid of $v\in[0,1]$, the diagonal difference $D(v)=C(v,v)-(C^{\top}\ast C)(v,v)$. If $C$ is stochastically increasing and $D(v)<0$ for even one $v$, then $\xi(C)>\psi(C)$ and the lower boundary of $R_{\mathrm{SI}}$ fails; if for any copula $\psi(C)>\sqrt{\xi(C)}$, the upper-bound theorem fails. The theorems predict no such point exists and that the Fréchet copula is the unique maximizer at each level, so a single numerical counterexample would settle the claim.
Extended reading notes
Core claim
The paper's central claim is Theorem 2.4: the attainable $(\xi,\psi)$-region for stochastically increasing copulas is exactly $R_{\mathrm{SI}}=\{(x,y)\in[0,1]^2\mid x\le y\le \sqrt{x}\}$. The lower edge $x=y$ is attained by ordinal sums of the independence copula, and the upper edge $y=\sqrt{x}$ is traced uniquely by the Fréchet copula family $C^{\mathrm{Fr}}_\alpha=(1-\alpha)\Pi+\alpha M$, $\alpha\in[0,1]$. This upper edge is not special to the SI class: Theorem 2.1 states that for every copula $C$, $\psi(C)\le\sqrt{\xi(C)}$, and for fixed $\xi=x$ the only maximizer is the Fréchet copula with $\alpha=\sqrt{x}$. The paper also characterizes equality $\xi=\psi$ inside the SI class by a two-point form of the conditional density, and for the full copula region it proves convexity and closedness, gives a Jensen-derived lower boundary curve, and shows that at $\psi=-1/2$ the smallest possible $\xi$ is $1/2$, attained only by the $2\times2$ checkerboard copula with mass on the off-diagonal squares.
Load-bearing premise
The load-bearing premise is the known diagonal inequality $(C^{\top}\ast C)(v,v)\le C(v,v)$ for stochastically increasing copulas; the full-region lower bound additionally rests on a rearrangement lemma from the author's companion preprint, so either failure would collapse the corresponding boundary.
Editorial extensions
If this is right
- Knowing $\xi$ for an SI copula determines $\psi$ up to the band $\xi\le\psi\le\sqrt{\xi}$, and both endpoints of the band are attainable at every level.
- For any copula whatsoever, $\psi\le\sqrt{\xi}$ with a unique extremizer at each level: only the Fréchet copula with parameter $\sqrt{x}$ reaches $\psi=\sqrt{x}$ when $\xi=x$.
- Equality $\xi=\psi$ is possible at every value from 0 to 1, realized by ordinal sums of the independence copula, so the two measures coincide on a full diagonal of structures.
- The corollary $\xi\le\frac{3}{4}\tau+\frac{1}{4}$ for SI copulas connects the square-root band to Kendall's tau.
- The full attainable region is convex and closed, so mixtures of copulas with the same $\psi$ fill out intervals of possible $\xi$ values between the two endpoint copulas.
Reading between the lines
- The paper leaves implicit that the upper ceiling $\psi\le\sqrt{\xi}$ is class-independent: any convex family of copulas containing the Fréchet family will share the same square-root upper boundary, so only the lower boundary distinguishes one subclass from another.
- Because the Jensen lower-bound curve is generated by functions that are not true copulas, the exact lower boundary of the full region remains open; a concrete next target is to identify whether the two-parameter family described in Section 3.2, or some refinement of it, actually attains it.
- Proposition 2.2 suggests a quantitative stability version: for an SI copula, the gap $\xi-\psi$ should measure the distance from the two-point conditional-density form, which could be turned into a diagnostic for near-functional dependence.
- Remark 2.6(c) identifies the attainable region for lower semilinear copulas with the SI region; if the same convexity argument applies to other subclasses that contain the Fréchet family and are closed under mixtures, exact regions for those classes would follow from their equality cases alone.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies the attainable joint range of Chatterjee's rank correlation ξ and Spearman's footrule ψ over bivariate copulas. The main results are: (i) Theorem 2.1 proves that for each x∈[0,1] the maximum possible ψ given ξ=x is √x, uniquely attained by the Fréchet copula C^Fr_α with α=√x; the proof uses a convex optimization reformulation over partial derivatives and verifies the KKT conditions as well as a quadratic-growth second-order condition. (ii) Theorem 2.4 characterizes the exact (ξ,ψ)-region for stochastically increasing copulas as R_SI = {(x,y): x≤y≤√x}, with the lower boundary attained by ordinal sums of the independence copula and the upper boundary uniquely by the Fréchet family. (iii) Section 3 provides a lower bound for the minimal ψ given ξ, based on Jensen's inequality, proves that at ψ=-1/2 the minimal ξ is 1/2 (uniquely attained by a checkerboard copula), and constructs a two-parameter copula family that approximately realizes the lower boundary. The paper is clearly written and explicitly acknowledges that the lower boundary for all copulas is not exactly characterized.
Significance. If correct, the upper-bound theorem and the exact SI region are substantive, elegant results: they give a sharp, simple geometric description of the joint range of two popular dependence measures, and the uniqueness of the Fréchet maximizer is non-obvious. The KKT proof of Theorem 2.1 is detailed and rigorous, including stationarity, primal/dual feasibility, complementarity, positive-definite Hessian, and convexity-based global minimality. The paper also gives a clean construction (ordinal sums of Π) attaining the full diagonal in the SI region, and it is honest about the fact that the global lower boundary is only bounded, not exactly reached. A notable caveat is that the SI-region result rests on the external pointwise inequality (C^T*C)(v,v) ≤ C(v,v) for stochastically increasing copulas, cited from [24, Thm. 4.2]; the manuscript does not re-derive this inequality, so the exactness of R_SI is contingent on that published theorem. Additionally, several supporting results in Section 3 are quoted from the author's own preprints [4] and [23], which are not yet peer-reviewed.
major comments (3)
- [Section 3.3, proof of Theorem 3.3 (convexity of R)] The proof that the full attainable region R is convex is incomplete. The argument mixes two lower-boundary copulas C0 and C1 and observes that x_λ := ξ(C_λ) ≤ (1−λ)x0+λx1 and y_λ := ψ(C_λ) = (1−λ)φ(x0)+λφ(x1). However, membership of (x_λ,y_λ) in R gives only y_λ ≥ φ(x_λ), and since x_λ may be strictly smaller than the weighted average, this does not imply the convexity inequality φ((1−λ)x0+λx1) ≤ (1−λ)φ(x0)+λφ(x1). The subsequent sentence that the point 'lies weakly below the straight line' does not establish that it lies above the lower boundary. Moreover, the assertion that 'R is only limited by the upper and lower boundaries' requires a proof that vertical slices are intervals; the paper only establishes that horizontal slices R_y are intervals. The closure proof depends on the same convexity claim. Please provide a complete proof or weaken the statement of Theorem 3.3 by omitting the convexity/closedness assertion, which is not needed for the inclusion (25).
- [Theorem 3.3, statement (25)] The displayed inclusion in Theorem 3.3 is not well-defined for y>0. The parameter µ(y) is introduced as the unique solution in [0,2] to the cubic equation, but as the proof itself notes, the equation has such a solution only for y∈[−1/2,0], because ψ(C↘_µ) ranges from 0 to −1/2 on that interval. For y>0, the expression ξ(C↘_µ(y)) is therefore undefined, yet the set in (25) ranges over y∈[−1/2,1]. Please restrict the lower-bound inequality to y≤0 and state separately that for y>0 the only bound used is y≤√x (equivalently x≥y^2).
- [Section 3.3, reduction to SD copulas and external dependencies] The proof of Theorem 3.3 relies on [4, Prop. 2.4 & Cor. 2.5] (the existence of an SD rearrangement preserving ξ and not increasing ψ) and on [2, Cor. 3.6] (continuity of ξ on C_SD). Both are results from unpublished preprints by the author and coauthors. Since this reduction is load-bearing for the lower-bound inclusion (25), the manuscript should either state these results explicitly with proofs or cite published versions; otherwise a reader cannot verify the key step that inf{ψ(C): ξ(C)=x} equals the infimum restricted to SD copulas. If the results are correct, this is a strength, but the current manuscript makes the verification difficult.
minor comments (4)
- [Section 3.2, Eq. (33)] The formula α(µ) = 1/2 − 2/5 µ for µ>2 yields negative values (e.g., µ=2.5 gives α=−0.5), contradicting the standing assumption α∈[0,0.5). This makes the path C_µ undefined for µ>2. If the intention was to continue from the point (0.3,0.5) toward (0.5,0.5) along β=0.5, a different expression (possibly 1/2 − 2/(5µ)) is needed.
- [Proof of Theorem 3.2, pointwise problem (22)] The displayed objective 'minimize µvh 1 + vh 2 1 + (1−v)h 2 2' is typographically confusing; it should read µ v h_1 + v h_1^2 + (1−v) h_2^2. Please clarify the notation.
- [Introduction and references] Several load-bearing results are cited from the author's preprints [4] and [23], as well as from other very recent preprints ([1], [2], [14]). It would help the reader if the introduction or a remark explicitly stated which external results are required for each main theorem and which of these are not yet published.
- [Figure 1 caption] The caption refers to a 'solid line from Π to C#' and a 'dotted line from Π to C#' as lower-bound curves. In black-and-white print these two curves may be difficult to distinguish; please use more distinct line styles or colors and add explicit labels in the figure itself.
Circularity Check
No construction-level circularity: the central upper-bound and SI-region results come from explicit optimization and external theorems; only minor self-citations appear in auxiliary lemmas.
full rationale
The derivation chain is not circular. The headline upper bound (Theorem 2.1) is obtained by solving a convex optimization problem in the partial derivative h, with objective determined by ψ and constraint ξ≤x; the candidate h* is explicitly checked against the KKT conditions and is exactly the derivative of the Fréchet copula with α=√x. No parameter is fitted to the target and then renamed as a prediction; the equality ψ=√ξ follows from the explicit optimizer. The SI region (Theorem 2.4) uses the external Markov-product inequality (C^T*C)(v,v)≤C(v,v) from [24, Thm. 4.2] and [24, Thm. 5.1] for idempotence of ordinal sums of Π; these are genuine external supports, not self-citations. The diagonal x=y is attained by ordinal sums of Π and the interior is filled by convexity plus continuity of ξ along the explicit quadratic path (15), so the region is not assumed into existence. The lower-boundary part is also handled honestly: C↘_μ is explicitly said not to be a true copula, and the α=3/5β choice in Section 3.2 is explicitly described as numerically motivated and not claimed as a derived law, so it does not disguise a fit as a prediction. The only mild concern is that the optimization framework, Lemma 1.1, and a few auxiliary facts are cited from the author’s own preprint [4], and Theorem 3.4’s key inequality at ψ=−1/2 is imported from the same author’s preprint [23] together with external [13, Thm. 3.2] and [1, Thm. 2.2 & Cor. 2.3]. These are technical lemmas rather than the target result, and the central ξ≤ψ≤√ξ characterization does not reduce to them by construction. Accordingly, the paper has no significant circularity; the score of 2 reflects minor self-citations that are not load-bearing for the main claim.
Assumptions & free parameters
free parameters (1)
- Heuristic α/β ratio in C_µ path =
α(µ)=3/20 µ for µ≤2, β(µ)=µ/4
assumptions (6)
- domain assumption Lemma 1.1 characterization of copulas via partial derivatives (from [4])
- domain assumption [24, Thm. 4.2]: for SI copulas, (C⊤∗C)(v,v) ≤ C(v,v) for all v
- domain assumption [24, Thm. 5.1]: ordinal sums of Π are SI, symmetric, and idempotent
- ad hoc to paper [4, Prop. 2.4 & Cor. 2.5]: every copula admits an SD rearrangement with ψ(C′)≤ψ(C) and ξ(C′)=ξ(C)
- standard math [5, Thm. 3.63]: quadratic growth condition for KKT sufficiency in Banach spaces
- domain assumption [13, Thm. 3.2] and [23, Thm. 4.1]: characterize ψ=-0.5 copulas as 2×2 off-diagonal checkerboards and give lower bound on ξ for checkerboards
Cite this review
Pith. "Pith review of On the exact region between Chatterjee's rank correlation and Spearman's footrule." pith.science (2026). https://pith.science/paper/6S6YPYQP
@misc{pith2026250907232,
author = {Pith},
title = {Pith review of: On the exact region between Chatterjee's rank correlation and Spearman's footrule},
year = {2026},
howpublished = {\url{https://pith.science/paper/6S6YPYQP}},
note = {Machine review of arXiv:2509.07232}
}
abstract
Chatterjee's rank correlation \(\xi\) has emerged as a popular measure quantifying the strength of directed functional dependence between random variables $X$ and $Y$. If $X$ and $Y$ are continuous, $\xi$ equals Spearman's footrule~\(\psi\) for the Markov product of the copula induced by $(X,Y)$ and its transpose. We analyze the relationship between these two measures more in depth by studying the attainable region of possible pairs \((\xi, \psi)\) over all bivariate copulas. In particular, we show that for given $\xi$, the maximal possible value of $\psi$ is uniquely attained by a Fr\'echet copula. As a by-product of this and a known result for Markov products of copulas, we obtain that \(\xi\le\psi\le \sqrt{\xi}\) characterizes the exact region of stochastically increasing copulas. Regarding the minimal possible value of \(\psi\) for given \(\xi\), we give a lower bound based on Jensen's inequality and construct a two-parameter copula family that comes comparably close.
Figures
Reference graph
Works this paper leans on
-
[24]
Stochastic monotonicity and the Markov product for copulas.J
Karl Friedrich Siburg and Christopher Strothmann. Stochastic monotonicity and the Markov product for copulas.J. Math. Anal. Appl., 503(2):14, 2021
work page 2021
-
[4]
The exact region and an inequality between Chatterjee's and Spearman's rank correlations
Jonathan Ansari and Marcus Rockel. The exact region and an inequality between chatterjee’s and spearman’s rank correlations.arXiv preprint arXiv:2506.15897, 2025
work page Pith review arXiv 2025
-
[13]
On the lower bound of Spearman’s footrule.Depend
Sebastian Fuchs and Yann Mccord. On the lower bound of Spearman’s footrule.Depend. Model., 7:126–132, 2019
work page 2019
-
[23]
Measures of association for approximating copulas.arXiv preprint arXiv:2505.08045, 2025
Marcus Rockel. Measures of association for approximating copulas.arXiv preprint arXiv:2505.08045, 2025
arXiv 2025
-
[1]
Jonathan Ansari and Sebastian Fuchs. A simple extension of azadkia and chatterjee’s rank correlation to a vector of endogenous variables.arXiv preprint arXiv:2212.01621, 2022
arXiv 2022
-
[2]
Jonathan Ansari and Sebastian Fuchs. On continuity of chatterjee’s rank correlation and related dependence measures.arXiv preprint arXiv:2503.11390, 2025
arXiv 2025
-
[3]
Dependence properties of bivariate copula families
Jonathan Ansari and Marcus Rockel. Dependence properties of bivariate copula families. Depend. Model., 12:36, 2024
2024
-
[5]
Fr´ ed´ eric Bonnans and Alexander Shapiro.Perturbation Analysis of Optimization Problems
J. Fr´ ed´ eric Bonnans and Alexander Shapiro.Perturbation Analysis of Optimization Problems. Springer Ser. Oper. Res. New York, NY: Springer, 2000. 16
work page 2000
Show all 27 references
-
[6]
On the exact region determined by Spearman’s footrule and Gini’s gamma.J
Damjana Kokol Bukovˇ sek and Blaˇ z Mojˇ skerc. On the exact region determined by Spearman’s footrule and Gini’s gamma.J. Comput. Appl. Math., 410:13, 2022. Id/No 114212
2022
-
[7]
A new coefficient of correlation.J
Sourav Chatterjee. A new coefficient of correlation.J. Amer. Statist. Assoc., 116(536):2009– 2022, 2021
2009
-
[8]
Darsow, Bao Nguyen, and Elwood T
William F. Darsow, Bao Nguyen, and Elwood T. Olsen. Copulas and Markov processes.Ill. J. Math., 36(4):600–642
-
[9]
Siburg, and Pavel A
Holger Dette, Karl F. Siburg, and Pavel A. Stoimenov. A copula-based non-parametric mea- sure of regression dependence.Scand. J. Stat., 40(1):21–41, 2013
2013
-
[10]
On the approximation of copulas via shuffles of Min.Statist
Fabrizio Durante and Juan Fern´ andez S´ anchez. On the approximation of copulas via shuffles of Min.Statist. Probab. Lett., 82(10):1761–1767, 2012
2012
-
[11]
Boca Raton, FL: CRC Press, 2016
Fabrizio Durante and Carlo Sempi.Principles of Copula Theory. Boca Raton, FL: CRC Press, 2016
2016
-
[12]
Quantifying directed dependence via dimension reduction.J
Sebastian Fuchs. Quantifying directed dependence via dimension reduction.J. Multivariate Anal., 201:21, 2024
2024
-
[14]
On exact regions between measures of concordance and chatterjee’s rank correlation for lower semilinear copulas.arXiv preprint arXiv:2507.23316, 2025
Sebastian Fuchs, Carsten Limbach, and Fabian Sch¨ urrer. On exact regions between measures of concordance and chatterjee’s rank correlation for lower semilinear copulas.arXiv preprint arXiv:2507.23316, 2025
2025 arXiv
-
[15]
Spearman’s footrule and Gini’s gamma: a review with complements.J
Christian Genest, Johanna Neˇ slehov´ a, and Noomen Ben Ghorbal. Spearman’s footrule and Gini’s gamma: a review with complements.J. Nonparametr. Stat., 22(8):937–954, 2010
2010
-
[16]
A modern treatment of the theory of functions of a real variable
Edwin Hewitt and Karl Stromberg.Real and abstract analysis. A modern treatment of the theory of functions of a real variable. 3rd printing, volume 25 ofGrad. Texts Math.Springer, Cham, 1975
1975
-
[17]
On the exact regions determined by Kendall’s tau and other concordance measures.Mediterr
Damjana Kokol Bukovˇ sek and Nik Stopar. On the exact regions determined by Kendall’s tau and other concordance measures.Mediterr. J. Math., 20(3):16, 2023
2023
-
[18]
On the exact region determined by Spearman’s rho and Spearman’s footrule.J
Damjana Kokol Bukovˇ sek and Nik Stopar. On the exact region determined by Spearman’s rho and Spearman’s footrule.J. Comput. Appl. Math., 437:23, 2024
2024
-
[19]
Spearman’s footrule and Gini’s gamma: local bounds for bivariate copulas and the exact region with respect to Blomqvist’s beta.J
Damjana Kokol Bukovˇ sek, Tomaˇ z Koˇ sir, Blaˇ z Mojˇ skerc, and Matjaˇ z Omladiˇ c. Spearman’s footrule and Gini’s gamma: local bounds for bivariate copulas and the exact region with respect to Blomqvist’s beta.J. Comput. Appl. Math., 390:24, 2021. Id/No 113385
2021
-
[20]
Linear algebra, Galois theory, representation theory, group extensions and Schur multiplier
Ramji Lal.Algebra 2. Linear algebra, Galois theory, representation theory, group extensions and Schur multiplier. Infosys Sci. Found. Ser. Singapore: Springer, 2017
2017
-
[21]
Nelsen.An Introduction to Copulas
Roger B. Nelsen.An Introduction to Copulas. 2nd ed.New York, NY: Springer, 2006
2006
-
[22]
Peˇ cari´ c, Frank Proschan, and Y
Josip E. Peˇ cari´ c, Frank Proschan, and Y. L. Tong.Convex functions, partial orderings, and statistical applications, volume 187 ofMath. Sci. Eng.Boston, MA etc.: Academic Press, 1992
1992
-
[25]
M. Sklar. Fonctions de r´ epartition ` andimensions et leur marges.Publ. Inst. Stat. Univ. Paris, 8:229–231, 1960
1960
-
[26]
On a strong metric on the space of copulas and its induced dependence measure.J
Wolfgang Trutschnig. On a strong metric on the space of copulas and its induced dependence measure.J. Math. Anal. Appl., 384(2):690–705, 2011
2011
-
[27]
Revisiting the region deter- mined by spearman’sρand spearman’s footruleϕ.Journal of Computational and Applied Mathematics, 457:116259, 2025
Marco Tschimpke, Manuela Schreyer, and Wolfgang Trutschnig. Revisiting the region deter- mined by spearman’sρand spearman’s footruleϕ.Journal of Computational and Applied Mathematics, 457:116259, 2025. 17
2025
Reviewed August 15, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.