Some results on the convergence of (quasi-) copulas

doi:10.1016/j.fss.2011.06.013

Fuzzy Sets and Systems

Volume 191, 16 March 2012, Pages 113-121

https://doi.org/10.1016/j.fss.2011.06.013 Get rights and content

Abstract

It is shown that pointwise convergence of a sequence $(A_{n})_{n \in N}$ of copulas to a copula A is equivalent (1) to the convergence of the corresponding endographs and (2) to the convergence of the corresponding upper (or lower) $α$ -levels for all but at most countably many $α$ in [0,1] (all with respect to the Hausdorff metric). Examples are given that show that the countably many exceptions in (2) cannot be omitted. It is furthermore shown that the main results also hold on the bigger class of quasi-copulas.

Highlights

► Pointwise convergence of copulas is equivalent to convergence of their endographs. ► Convergence of copulas is equivalent to convergence of almost all upper level sets. ► Convergence of copulas is equivalent to convergence of almost all lower level sets. ► The convergence results also hold for the class of quasi-copulas.

References (12)

F. Durante et al.
Semilinear copulas
Fuzzy Sets Syst.
(2008)
G.A. Fredricks et al.
Copulas with fractal supports
Insur. Math. Econ.
(2005)
R.B. Nelsen et al.
Quasi-copulas and signed measures
Fuzzy Sets Syst.
(2010)
R.B. Nelsen et al.
Kendall distribution functions
Stat. Probab. Lett.
(2003)
W. Trutschnig
Characterization of the sendograph-convergence of fuzzy sets by means of their $L_{p}$ - and levelwise convergence
Fuzzy Sets Syst.
(2010)
M.F. Barnsley
Fractals Everywhere
(1993)

There are more references available in the full text version of this article.

Cited by (12)

A hitchhiker's guide to quasi-copulas
2020, Fuzzy Sets and Systems
This paper presents a review of the concept of an n-quasi-copula, starting from the characterization of functions that can be derived from operations on random variables. We recall several other characterizations and properties that have been proven in the literature, highlighting the differences between the bivariate case and the higher-dimensional case. Additionally, we recall several applications of n-quasi-copulas such as their role in the study of bounds on sets of n-copulas. We also recall various classes of n-quasi-copulas that have been studied in the literature. Finally, we present some questions that have not yet been answered in the literature.
A metric space of subcopulas — An approach via Hausdorff distance
2020, Fuzzy Sets and Systems
On the length of copula level curves
2018, Journal of Multivariate Analysis
Motivated by the well-known fact that the surface of copulas is closely related to common dependence measures such as Spearman’s rho, we investigate level curves of bivariate copulas and study their lengths. To this end, we establish the length profile $L_{C} (t)$ which maps each level $t \in [0, 1]$ to the length of the respective level curve. Some basic properties of the length profile, such as continuity and differentiability with respect to $t$ , are examined. Based on the length profile, a measure $ℓ_{C}$ is defined, which can be interpreted as the average level curve length. $ℓ_{C}$ is a measure of association, it is, however, not a concordance measure in general. Some further, partially surprising properties, such as closed-form formulas of $ℓ_{C}$ for completely dependent copulas, conclude the paper.
Extensions of subcopulas
2017, Journal of Mathematical Analysis and Applications
In view of Sklar's Theorem the probability distribution function of every (not necessarily continuous) random vector can be uniquely decomposed in terms of the marginal distributions of its components and a suitable subcopula. The study of such latter functions is therefore of interest for understanding the dependence information of non-continuous variables. Here, we investigate some analytical properties of the class of subcopulas, including compactness (with respect to a novel metric), approximations and Baire category results. Moreover, under a suitable assumption, we describe all possible extensions from a subcopula to a copula in any dimension.
Idempotent and multivariate copulas with fractal support
2012, Journal of Statistical Planning and Inference
Using special iterated function systems (IFS) Fredricks et al. (2005) constructed two-dimensional copulas with fractal supports and showed that for every $s \in (1,2)$ there exists a copula A whose support has Hausdorff dimension s. In the current paper we present a stronger version and prove that the same result holds for the subclass of idempotent copulas. Additionally we show that every doubly stochastic idempotent matrix N (having neither minimum nor maximum rank) induces a family of idempotent copulas such that, firstly, the corresponding Markov kernels transform according to N and, secondly, the set of Hausdorff dimensions of the supports of elements of the family covers (1,2). Furthermore we generalize the IFS approach to arbitrary dimensions $d \geq 2$ and show that for every $s \in (1, d)$ we can find a d-dimensional copula whose support has Hausdorff dimension s.
On the approximation of copulas via shuffles of Min
2012, Statistics and Probability Letters

View all citing articles on Scopus

View full text

Some results on the convergence of (quasi-) copulas

Abstract

Highlights

Fuzzy Sets Syst.

Insur. Math. Econ.

Fuzzy Sets Syst.

Stat. Probab. Lett.

Fuzzy Sets Syst.

Fractals Everywhere