Pith. sign in

REVIEW 3 major objections 5 minor 21 references

Directional $\rho$-coefficients

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proves that every n-dimensional directional rho coefficient equals a signed linear combination of lower-dimensional rho-minus coefficients of variable subsets, resolving the trivariate conjecture and correcting an earlier…

desk verdict Real population-level extension of directional rho coefficients to arbitrary dimensions, but the finite-sample estimator decomposition in Theorem 6 is only asymptotic, not exact — fixable and worth a serious referee. read the letter →

arxiv 2505.22206 v1 pith:CLUAUYP6 submitted 2025-05-28 math.ST stat.TH

Enrique de Amo , David García-Fernández , Manuel Úbeda-Flores This is my paper · ORCID
classification math.STstat.TH MSC 62H0562H12
keywords copuladirectionalrho-coefficientsmultivariatedependencenonparametricestimatorsSpearman'srhorank-basedestimationinclusion-exclusion
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Directional rho-coefficients, previously defined only for three variables, quantify how strongly a random vector deviates from independence along a chosen direction in which each variable counts either upward or downward. This paper generalizes the definition to arbitrary dimension and proves a representation theorem: for any direction, the n-dimensional coefficient is exactly a signed linear combination of lower-dimensional rho-minus coefficients of subsets of the variables. If the theorem is right, the open conjecture from the original trivariate paper is resolved, the trivariate formula in a 2013 paper is corrected, and every directional coefficient is computable from subset coefficients alone. The authors also give rank-based estimators that satisfy the identical linear decomposition and are asymptotically normal and consistent, with Monte Carlo support for Clayton copulas.

What carries the argument

The load-bearing object is the inclusion-exclusion expansion of the directional survival probability. For a direction that sets coordinates in $I$ downward and coordinates in $J$ upward, the joint probability $P\left(\bigcap_{i\in I}\{X_i<x_i\}\cap\bigcap_{i\in J}\{X_i>x_i\}\right)$ is expanded as an alternating sum of copulas $C(x_{I\cup S},1)$ over subsets $S\subseteq J$; this identity is quoted from Lemma 3.4 and Remark 3.5 of the authors' 2025 paper and is not reproduced in this manuscript. Substituting that expansion into Definition (2) and rewriting each integral in terms of the lower-dimensional coefficient $\rho^-_{X_{I\cup S}}$ yields Theorem 4. The same expansion, applied to rank-based pseudo-observations $U_{ij}=R_{ij}/(n+1)$ with directional ranks, gives the linear decomposition of the estimator in Theorem 6; weak convergence of the underlying empirical copula process (Theorems 7 and 8) then yields Corollary 9.

What would settle it

Numerically integrate Definition (2) for a fixed 4-copula, for instance a Clayton copula with theta = 2, in the direction (-1,1,1,-1); separately compute the rho-minus values of every subset {i} union S from the corresponding lower-dimensional margins and combine them by Theorem 4 or Table 2. Exact numerical agreement is required, and any mismatch would falsify the representation.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is Theorem 4: for an n-dimensional random vector with copula $C$ and any direction $\alpha\in\{-1,1\}^n$, the directional coefficient $\rho^\alpha_n(C)$ introduced in Definition (2) equals a signed linear combination of lower-dimensional rho-minus coefficients $\rho^-_{X_{I\cup S}}$ of the subsets $X_{I\cup S}$, where $I$ is the set of coordinates with $\alpha_i=-1$ and $S$ ranges over subsets of the remaining coordinates. The combination weights depend only on the sizes $|I|$ and $|S|$, not on the copula itself. This result resolves the conjecture from the 2011 trivariate paper and corrects Equation (10) of the 2013 trivariate paper. The rank-based estimator $\hat\rho^\alpha_d$ satisfies the same linear decomposition (Theorem 6), and under a differentiability condition on the copula and a tightness condition on the associated empirical process it is asymptotically normal and consistent (Corollary 9). The paper reads the result as a tool for detecting directional dependence that pairwise measures miss, and illustrates it on daily log-returns of three stocks.

Load-bearing premise

The proof of Theorem 4 relies entirely on an inclusion-exclusion identity (Lemma 3.4 and Remark 3.5 of the authors' 2025 paper) that is quoted but neither stated nor proved here; if that identity fails for some dimension, the representation theorem and the estimator decomposition built on it have no foundation.

Editorial extensions

If this is right

  • In any dimension, all directional coefficients are determined by the lower-dimensional rho-minus values of the marginal subsets, so no new high-dimensional integral is needed once the subset coefficients are known.
  • Rank-based estimators inherit the same linear structure, so estimating subset rho-minus values directly yields an estimate of any directional coefficient.
  • Corollary 9 gives asymptotic normality and consistency of the directional estimators under the stated conditions, enabling confidence intervals and tests for directional dependence.
  • The trivariate case coincides with the existing index of [7], while Equation (10) of that paper is corrected.
  • Monte Carlo simulations for Clayton copulas in dimensions 3 and 4 show the estimators approaching the coefficients as sample size grows.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The decomposition identifies all directional dependence information with lower-dimensional rho-minus concordance, so an analyst could select the direction maximizing the absolute coefficient by estimating only subset rho-minus values, a potentially large computational saving in high dimensions.
  • The same linear structure may transfer to other orthant-based concordance measures, since the argument uses only the inclusion-exclusion structure of orthant probabilities.
  • Because Theorem 4 expresses the directional coefficient through subset margins, it suggests a diagnostic for high-dimensional copula selection: a fitted copula whose subset rho-minus values are matched should reproduce the directional coefficients.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper extends the directional rho-coefficients of Nelsen and Ubeda-Flores to arbitrary dimension, proves a representation theorem (Theorem 4) expressing every directional coefficient as a linear combination of lower-dimensional rho-minus coefficients, and claims to correct Equation (10) of Garcia, Gonzalez-Lopez, and Nelsen. It also proposes rank-based estimators, states an exact finite-sample linear decomposition of these estimators (Theorem 6), and derives asymptotic normality and consistency (Corollary 9). The paper includes Monte Carlo simulations for Clayton and FGM copulas and an application to daily stock returns.

Significance. If Theorem 4 is correct, it resolves the conjecture in [14] and provides a principled way to compute directional coefficients in arbitrary dimension from lower-dimensional components. The proof of Theorem 4 is essentially verifiable from the displayed inclusion-exclusion expansion, and the closed-form examples for the minimum copula and the FGM family are checkable and appear correct. The central idea is useful and the asymptotic estimator is natural. However, the claimed exact finite-sample decomposition in Theorem 6 is false as stated, Example 4 reports Monte Carlo signs opposite to the closed-form FGM formula, and Eq. (8) has an algebraic inconsistency. These issues are fixable but currently undermine the estimator section, so the paper needs major revision before it is publishable in its present form.

major comments (3)
  1. [Section 4, Theorem 6, Eq. (4)] The claimed exact linear decomposition of the finite-sample estimator is algebraically false when the symbols b-rho-minus refer to the estimators defined in Section 4. For n=3, d=2, alpha=(1,-1), and perfect positive dependence (R_{1j}=R_{2j}=j for all j), the direct estimator equals (10/3-4)/(14/3-4)=-1, while Eq. (4) with the Section 4 estimator b-rho^-_{12}=1 gives (4^2/(2/3))*(-1/12)=-2. The identity becomes correct only in the limit n tends to infinity. The source of the discrepancy is that the proof of Theorem 6 defines b-rho^-_K with the normalization 2^m(m+1)/(2^m-(m+1)) applied to (1/n) sum prod(1-U) - 2^{-m}, which is not the Section 4 estimator b-rho^-_m = [ (1/n) sum prod R - ((n+1)/2)^m ] / [ (1/n) sum j^m - ((n+1)/2)^m ]. Theorem 6 should therefore be restated as an asymptotic equivalence, or all lower-dimensional estimators in Eq. (4) must be redefined to match the proof's construction.
  2. [Section 4, Example 4] The Monte Carlo signs reported in Example 4 contradict the closed-form expression in Example 3. For the FGM 3-copula with lambda=0.6, Example 3 gives rho^-_3 = +lambda/27 = +0.0222 and rho^+_3 = -0.0222, whereas Example 4 reports b-rho^-_3 = -0.0215 and b-rho^+_3 = +0.0217. Either the direction labels in the simulation are swapped or the reported values are erroneous. Since the example is used to illustrate the limitation of rho^*_3, this contradiction must be resolved before the numerical claims can be trusted.
  3. [Section 4.1.2, Eq. (8)] Equation (8) is not algebraically consistent with the definition of the empirical copula in Eq. (7). From (7), the integral over I^d of C_{beta,alpha,n} equals (1/(n+1)^{d+1}) sum_{j=1}^n prod_{i=1}^d R^{alpha_i}_{ij}, so substituting this into Eq. (8) does not reproduce the estimator's numerator (1/n) sum_j prod_i R^{alpha_i}_{ij} - ((n+1)/2)^d unless the prefactor contains an additional factor (n+1)/n. Because the proof of Corollary 9 relies on Eq. (8), the prefactor must be corrected and the asymptotic normality proof adjusted accordingly.
minor comments (5)
  1. [Section 3, proof of Theorem 4] The proof begins by invoking [1, Lemma 3.4 and Remark 3.5], an unstated result from a paper by three of the four authors. The displayed inclusion-exclusion expansion can be verified directly, so the theorem is not compromised, but the paper would be more self-contained and easier to check if that lemma were stated here or if the expansion were derived without the unexplained citation.
  2. [Sections 4 and Theorem 6] The symbol n is used both for the sample size and for the dimension of the random vector; for example, Theorem 6 states alpha in R^n but the vector has dimension d and the sample size is n. This notational collision is confusing and should be fixed, for instance by using m for sample size.
  3. [Section 3, Theorem 4] Theorem 4 contains terms of the form rho^-_{X_I} or rho^-_{X_{I cup S}} for subsets of size one (and, when I is empty, size zero), for which rho^- is not defined by Eq. (2) because the normalization vanishes. These terms carry zero coefficients, so they should be explicitly omitted from the summation or accompanied by a convention such as assigning them the value zero.
  4. [Section 4, Table 6] The text says the values in Table 6 were obtained using the expression in Theorem 6, but the numerical results match the asymptotic coefficients 20/33, -10/11, and 1 from Theorem 4/Table 2 rather than the finite-sample prefactors in Eq. (4). For example, with n=20 and theta=0.4, substituting the displayed values of b-rho^-_{14}, b-rho^-_{124}, b-rho^-_{134}, and b-rho^-_{1234} into Eq. (4) gives approximately -0.072, not -0.0650. Please clarify which formula was actually used and correct the caption or the table accordingly.
  5. [Abstract and Introduction] There are several grammatical slips, such as 'correcting a result in [7] an erratum in the current literature' in the abstract and 'Jodgeo' in the Introduction (should be Jogdeo). These should be corrected in a final polish.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the cited lemma in Theorem 4 is an independent inclusion-exclusion identity, though Theorem 6 has a finite-sample correctness concern.

full rationale

The population derivation in Theorem 4 is not circular. Its proof opens with the self-cited line “By using [1, Lemma 3.4 and Remark 3.5]”, but the very next displayed equations spell out the inclusion-exclusion expansion of the joint event into copula terms C(x_{I∪S},1). That cited lemma is a parameter-free, standard identity about probabilities and copulas, not a restatement of the directional-rho representation being proved; the remaining argument is explicit integral algebra that cancels the constant terms and leaves the stated linear combination. Thus the central claim has independent mathematical content, and the self-citation does not reduce the theorem to its inputs. The estimator section is also not circular in the definitional sense: the estimators are rank-based plug-ins, and Corollary 9 uses a standard empirical-process delta method with explicit moment checks rather than fitting anything to the target coefficient. The most serious concern is a correctness issue rather than a circularity issue: in Theorem 6 the proof replaces the Section 4 definition of ρ̂^-ᵋ, whose denominator is n^{−1}Σjᵐ − ((n+1)/2)ᵐ, with the asymptotically normalized expression 2ᵐ(m+1)/(2ᵐ−(m+1))[n^{−1}Σ∏(1−U) − 2^{−ᵐ}]; these coincide only as n→∞, so equation (4) appears not to be an exact finite-sample identity. That flaw, if confirmed, does not make the paper's derivation circular. Overall, the circularity score is low because the load-bearing mathematics is independent of the self-citation and no fitted parameter is renamed as a prediction.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new fitted parameters or invented entities: the copula families in the examples have standard parameters λ and θ that are inputs, not fitted values. The main assumptions are standard copula theory, a self-cited inclusion-exclusion lemma, and a smoothness condition on the copula needed for empirical process convergence.

assumptions (3)
  • standard math Sklar's theorem and the copula representation of joint distributions.
    The entire framework assumes the joint distribution is uniquely represented by its copula via Sklar's theorem, invoked in Section 2.1.
  • domain assumption [1, Lemma 3.4 and Remark 3.5]: inclusion-exclusion expansion of an orthant probability into copula terms.
    The proof of Theorem 4 in Section 3 rests on this cited lemma from the authors' own 2025 paper; it is not reproduced in this manuscript.
  • domain assumption Condition 1 in Section 4.1: first-order partial derivatives of C_{β,α} exist and are continuous on the interior.
    This smoothness condition is required for the weak convergence of the empirical copula process in Theorem 7 and hence for the asymptotic normality of the estimators in Corollary 9.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Directional $\rho$-coefficients." pith.science (2026). https://pith.science/paper/CLUAUYP6

@misc{pith2026250522206,
  author       = {Pith},
  title        = {Pith review of: Directional $\rho$-coefficients},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CLUAUYP6}},
  note         = {Machine review of arXiv:2505.22206}
}
abstract

In this paper we obtain advances for the concept of directional $\rho$-coefficients, originally defined for the trivariate case in [Nelsen, R.B., \'Ubeda-Flores, M. (2011). Directional dependence in multivariate distributions. Ann. Inst. Stat. Math 64, 677-685] by extending it to encompass arbitrary dimensions and directions in multivariate space. We provide a generalized definition and establish its fundamental properties. Moreover, we resolve a conjecture from the aforementioned work by proving a more general result applicable to any dimension, correcting a result in [Garc\'ia, J.E., Gonz\'alez-L\'opez, V.A., Nelsen, R.B. (2013). A new index to measure positive dependence in trivariate distributions. J. Multivariate Anal. 115, 481-495] an erratum in the current literature. Our findings contribute to a deeper understanding of multivariate dependence and association, offering novel tools for detecting directional dependencies in high-dimensional settings. Finally, we introduce nonparametric estimators, based on ranks, for estimating directional $\rho$-coefficients from a sample.

Figures

Figures reproduced from arXiv: 2505.22206 by the authors.

Figure 1
Figure 1. Scatterplots of log-return data. The largest values for the estimators are in the directions (1, 1, 1) and (−1, −1, −1), which can be 20 [PITH_FULL_IMAGE:figures/full_fig_p020_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 21 canonical work pages

  1. [1]

    de Amo, E., Garc ´ ıa-Fern´ andez, D., Quesada-Molina, J.J.,´Ubeda-Flores, M. (2025). Modeling direc- tional monotonicity with copulas.Iranian J. Fuzzy Syst.22, 135–146

    work page 2025

  2. [14]

    Nelsen, R.B., ´Ubeda-Flores, M. (2011). Directional dependence in multivariate distributions.Ann. Inst. Stat. Math64, 677–685

    work page 2011

  3. [2]

    de Amo, E., Rodr ´ ıguez-Gri˜ nolo, M.R.,´Ubeda-Flores, M. (2024). Directional dependence orders of random vectors.Mathematics12, article 419

    work page 2024

  4. [3]

    (2016).Principles of Copula Theory

    Durante, F., Sempi, C. (2016).Principles of Copula Theory. Chapman & Hall/CRC, Boca Raton

    work page 2016

  5. [4]

    Fermanian, J.D., Radulovic, D., Wegkamp, M. (2004). Weak convergence of empirical copula pro- cesses.Bernoulli10, 847–860

    work page 2004

  6. [5]

    Fisher, N.I. (1997). Copulas. In:Encyclopedia of statistical sciences, Vol. 1 (S. Kotz, C.B. Read, D.L. Banks, Eds.), Wiley, New York, pp. 159–163

    work page 1997

  7. [6]

    (1987).Seminar on Empirical Processes

    Gaenssler, P., Stute, W. (1987).Seminar on Empirical Processes. Springer Basel, Basel

    work page 1987

  8. [7]

    Garc ´ ıa, J.E., Gonz´ alez-L´ opez, V.A., Nelsen, R.B. (2013). A new index to measure positive depen- dence in trivariate distributions.J. Multivariate Anal.115, 481–495

    work page 2013

Show all 21 references
  1. [8]

    Jogdeo, K. (1982). Concepts of dependence. In:Encyclopedia of Statistical Sciences, Vol. 1 (S. Kotz, N.L. Johnson, Eds.), Wiley, New York, pp. 324–334

    1982

  2. [9]

    Joe, H. (1990). Multivariate concordance.J. Multivariate Anal.35, 12–30

    1990

  3. [10]

    (2014).Dependence modeling with copulas

    Joe, H. (2014).Dependence modeling with copulas. Chapman & Hall, New York

    2014

  4. [11]

    M¨ uller, A., Scarsini, M. (2006). Archimedean copulae and positive dependence.J. Multivariate Anal. 93, 434–445

    2006

  5. [12]

    Nelsen, R.B. (1996). Nonparametric measures of multivariate association. In:Distributions with Fixed Marginals and Related Topics, Vol. 28 (L. R¨ uschendorf, B. Schweizer, M.D. Taylor, Eds.), Institute of Mathematical Statistics, Hayward, 223–232

    1996

  6. [13]

    (2006).An Introduction to Copulas (2nd ed.)

    Nelsen, R.B. (2006).An Introduction to Copulas (2nd ed.). Springer, New York

    2006

  7. [15]

    (2025).Quantitative Risk Management Library, Version 0.4-35

    Pfaff, B., Hofert, M., McNeil, A., Ulmann, S. (2025).Quantitative Risk Management Library, Version 0.4-35. https://CRAN.R-project.org/package=QRMlib

    2025

  8. [16]

    P´ erez, A., Prieto-Alaiz, M. (2016). A note on nonparametric estimation of copula-based multivariate extensions of Spearman’s rho.Stat. Probab. Lett.112, 41–50

    2016

  9. [17]

    Quesada-Molina, J.J., ´Ubeda-Flores, M. (2012). Directional dependence of random vectors.Inf. Sci. 215, 67–74

    2012

  10. [18]

    Quesada-Molina, J.J., ´Ubeda-Flores, M. (2024). Monotonic random variables according to a direc- tion.Axioms13, 275

    2024

  11. [19]

    Sklar, A. (1959). Fonctions de r´ epartition ` andimensions et leurs marges.Publ. Inst. Statist. Univ. Paris8, 229–231. 22

    1959

  12. [20]

    ´Ubeda-Flores, M., Fern´ andez-S´ anchez, J. (2017). Sklar’s theorem: The cornerstone of the Theory of Copulas. In:Copulas and Dependence Models with Applications(M. ´Ubeda-Flores, E. de Amo Artero, F. Durante, J. Fern´ andez-S´ anchez, Eds.), Springer, Cham, pp. 241–258

    2017

  13. [21]

    Wei, Z., Wang, T., Panichkitkosolkul, W. (2014). Dependence and association concepts through copulas. In:Modeling Dependence in Econometrics - Advances in Intelligent Systems and Computing, Vol. 251 (V.N. Huynh, V. Kreinovich, S. Sriboonchitta, Eds.), Springer, Cham, pp. 113–126. 23

    2014

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.