On bivariate Archimax copulas: Level sets, mass distributions and related results
Nicolas Pascal Dietrich · 2025
Read On bivariate Archimax copulas: Level sets, mass distributions and related results by Nicolas Pascal Dietrich (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-17045.txt.
Introduction
Nicolas Dietrich studies how probability mass spreads inside a family of dependence models. His paper, from twenty twenty-five, treats bivariate Archimax copulas. Copulas join separate risk curves into a joint picture. Extreme value copulas are a famous subfamily, and their mass patterns were mapped in past work.
Dietrich extends that mapping to the wider Archimax family. Using conditional distributions, he shows the support rests on a few key functions. Any lump of point mass sits on the graphs of convex or rising functions. Flat level sets of the model get the same treatment. A classic distribution formula falls out as a byproduct, with simple level masses.
Catalog record
| CREATOR | Nicolas Pascal Dietrich |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 150 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-17045
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On bivariate Archimax copulas: Level sets, mass distributions and related results Nicolas Pascal Dietrich Abstract Motivated by the results in ( Mai and Scherer, 2011 ; Trutschnig et al., 2016 ) , which examine the way bivariate Extreme Value copulas distribute their mass, we extend these findings to the larger family of bivariate Archimax copulas 𝒞 a m \mathcal{C}_{am} . Working with Markov kernels (conditional distributions), we analyze the mass distributions of Archimax copulas C ∈ 𝒞 a m C\in\mathcal{C}_{am} and show that the support of C C is determined by some functions f 0 f^{0} , g L g^{L} and g R g^{R} . Additionally, we prove that the discrete component (if any) of C C concentrates its mass on the graphs of certain convex functions f s f^{s} or non-decreasing functions g t g^{t} . Investigating the level sets L t L^{t} of Archimax copulas C ∈ 𝒞 a m C\in\mathcal{C}_{am} , we establish that these sets can also be characterized in terms of the afore-mentioned functions f s f^{s} and g t g^{t} . Furthermore, recognizing the close relationship between the level sets L t L^{t} of a copula C C and its Kendall distribution function F C K F_{C}^{K} , we provide an alternative proof for the representation of F C K F_{C}^{K} for arbitrary Archimax copulas C ∈ 𝒞 a m C\in\mathcal{C}_{am} and derive simple expressions for the level set masses μ C ( L t ) \mu_{C}(L^{t}) . Building upon the fact that Archimax copulas C ∈ 𝒞 a m C\in\mathcal{C}_{am} can be represented via two univariate probability measures γ \gamma and ϑ \vartheta — so-called Williamson and Pickands dependence measures — we show that absolute continuity, discreteness and singularity properties of these measures γ \gamma and ϑ \vartheta carry over to the corresponding Archimax copula C γ , ϑ C_{\gamma,\vartheta} . Finally, we derive conditions on γ \gamma and ϑ \vartheta such that the support of the absolutely continuous, discrete or singular component of C γ , ϑ C_{\gamma,\vartheta} coincides with the support of C γ , ϑ C_{\gamma,\vartheta} . Keywords : Archimax copula, Extreme Value copula, Archimedean copula, Markov kernel, Pickands dependence function 2020 MSC : 62G32, 62H05, 26A30, 60A10, 60E05 1 Introduction Examining maxima of independent, identically distributed sequences of random variables, Extreme Value copulas (EVCs) arise naturally and are a well-suited tool for modeling dependence between extreme events. Including various parametric sub-families, such as Gumbel–Hougaard and Galambos copulas ( Gudendorf and Segers, 2011 ) , EVCs are applied in numerous fields, notably finance ( Longin and Solnik, 2001 ; McNeil et al., 2005 ) and hydrology ( Salvadori, 2007 ) . Moreover, EVCs are also theoretically appealing, since they are characterized by a single univariate function A . To be more precise, according to ( Durante and Sempi, 2015 ; Haan and Resnick, 1977 ; Nelsen, 2006 ; Pickands, 1981 ) a copula B is an EVC if, and only if there exists a Pickands dependence function A , i.e., a convex function A\colon[0,1]\rightarrow[\frac{1}{2},1] fulfilling \max\{1-x,x\}\leq A(x)\leq 1 for every x\in[0,1] such that the identity B(x,y)=\exp\left((\log(x)+\log(y))A\left(\frac{\log(x)}{\log(x)+\log(y)}\right)\right) holds for every x,y\in(0,1) . The family of EVCs has been the subject of extensive study, yielding key results on convergence, mass distributions, and other analytic and measure-theoretic properties ( Kasper et al., 2021 ; Mai and Scherer, 2011 ; Trutschnig et al., 2016 ) . Considering so-called Archimedean generators \psi (defined in Section 2.2.1 ) now allows to naturally extend the family of EVCs to the much larger family of Archimax copulas. Following ( Capéraà et al., 2000 ) , a copula C is called an Archimax copula if, and only if there exists a convex Archimedean generator \psi\colon[0,\infty)\rightarrow[0,1] with pseudo-inverse \varphi and a Pickands dependence function A\colon[0,1]\rightarrow[\frac{1}{2},1] such that C(x,y)=\psi\left((\varphi(x)+\varphi(y))A\left(\frac{\varphi(x)}{\varphi(x)+\varphi(y)}\right)\right) holds for every x,y\in(0,1) . It was proved in ( Capéraà et al., 2000 ) and later in ( Durante et al., 2018 ) that a function C , defined in the afore-mentioned way indeed constitutes a bona fide copula. As the name suggests, the family of Archimax copulas \mathcal{C}_{am} encompasses both the family of EVCs \mathcal{C}_{ev} and the family of Archimedean copulas \mathcal{C}_{ar} as sub-families and can therefore be seen as an extension of these two families of copulas. For more background on Archimedean copulas and their applications, we refer to ( Durante and Sempi, 2015 ; McNeil and Nešlehová, 2009 ; Nelsen, 2006 ; Kasper et al., 2024 ) . Viewing Archimax copulas from an application perspective, they offer the flexibility to model asymptotic dependence in extreme events while simultaneously capturing joint risks at moderate levels ( Chatelain et al., 2020 ) . For further research in the field of bivariate Archimax copulas we refer to ( Bacigál et al., 2011 ; Capéraà et al., 2000 ; Genest and Jaworski, 2021 ) . Moreover, an extension to the multivariate setting was considered in ( Charpentier et al., 2014 ; Mesiar and Jágr, 2013 ) . Regarding inference, an estimator for a multivariate, semi-parametric framework was provided in ( Chatelain et al., 2020 ) . While it has already been established how Archimedean copulas and EVCs distribute their mass ( McNeil and Nešlehová, 2009 ; Kasper et al., 2024 ; Mai and Scherer, 2011 ; Trutschnig et al., 2016 ) , and the level sets of Archimedean copulas have been investigated in ( McNeil and Nešlehová, 2009 ; Kasper et al., 2024 ) , comparable results for the larger family of Archimax copulas \mathcal{C}_{am} remain, to the best of the author’s knowledge, unstudied. The current paper addresses this gap by analyzing the mass distributions and level sets of bivariate Archimax copulas, providing several new insights and related findings. Motivated by the fact that the level sets of Archimedean copulas can be represented by convex functions ( Nelsen, 2006 ) , we prove that the level sets of Archimax copulas can similarly be characterized by two functions f^{s} and g^{t} (specified in Section 3 ) for some s,t\in[0,1) . Moreover, we derive conditions on the Pickands dependence function A , such that the level sets L^{t} of the corresponding C\in\mathcal{C}_{am} coincide with the graphs of functions f^{t} and show that these functions are convex. Building upon the fact that both Archimedean copulas and EVCs can be characterized by univariate probability measures, so-called Williamson measures \gamma and Pickands dependence measures \vartheta , respectively ( Dietrich and Trutschnig, 2025 ; Kasper et al., 2024 ; McNeil and Nešlehová, 2009 ; Pickands, 1981 ) , we demonstrate that Archimax copulas can similarly be defined via these measures. Taking this into consideration and providing an explicit representation of (a version of) the Markov kernel K_{C} of C\in\mathcal{C}_{am} , we study the support of C and show how C distributes its mass. To be more precise, we prove that if either of the measures \gamma or \vartheta has full support, the support of C_{\gamma,\vartheta}\in\mathcal{C}_{am} is fully characterized by the afore-mentioned functions f^{s} and g^{t} for some s,t\in[0,1) . Furthermore, we derive conditions on \gamma and \vartheta such that C_{\gamma,\vartheta}\in\mathcal{C}_{am} has full support and establish that the discrete component (as specified in Section 2.1 and if non-degenerated) of C_{\gamma,\vartheta} concentrates its mass