KYBERNETIKA — VOLUME 45 (2009), NUMBER 6, PAGES 992 – 1002

CONSTRUCTING COPULAS BY MEANS OF PAIRS

OF ORDER STATISTICS

Ali Dolati and Manuel Úbeda-Flores

In this paper, we introduce two transformations on a given copula to construct new

and recover already-existent families. The method is based on the choice of pairs of order
statistics of the marginal distributions. Properties of such transformations and their effects
on the dependence and symmetry structure of a copula are studied.

Keywords: copula, dependence ordering, FGM family, measure of association, symmetry,

transformation
AMS Subject Classification: 60E05, 62H20

1. INTRODUCTION
The construction of distributions with given marginals has been a problem of interest
to statisticians for many years. Today, in view of Sklar’s theorem [25, 26], this
problem can be reduced to the construction of a copula. Copulas are a special type
of aggregation functions, and nowadays are of interest in fuzzy set theory [5, 6, 12,
14, 16].
Nelsen [22] summarizes different methods of constructing copulas. Recently, vari-
ous authors provide construction methods from the class of copulas to itself, or from
a more general class of functions to another (see, for instance, [2, 8, 10, 11, 13, 18,
19, 20]). One of the purposes for such constructions is to increase the availability of
copulas for modeling purposes.
In this paper we provide two new transformations of copulas, which recover some
known families of copulas. The method is based on the choice of pairs of order statis-
tics of the marginal distributions. Properties on dependence, symmetry, invariance,
and random number generation of such transformations are also shown. Preliminary
ideas can be found in [3].

2. PRELIMINARIES
A (bivariate) copula is the restriction to [0, 1]2 of a continuous bivariate distribution
function whose margins are uniform on [0, 1]. The importance of copulas as a tool for
statistical analysis and modeling stems largely from the observation that the joint
Copulas Based on Pairs of Order Statistics 993

distribution H of the random vector (X, Y ) with respective margins F and G can
be expressed by H(x, y) = C(F (x), G(y)), (x, y) ∈ [−∞, ∞]2 , where C is a copula
that is uniquely determined on Range F ×Range G (Sklar’s theorem).
Equivalently, a copula is a function C : [0, 1]2 −→ [0, 1] which satisfies:

(C1) the boundary conditions C(t, 0) = C(0, t) = 0 and C(t, 1) = C(1, t) = t for all
t ∈ [0, 1],

(C2) the 2-increasing property, i. e., C(x2 , y2 )−C(x2 , y1 )−C(x1 , y2 )+C(x1 , y1 ) ≥ 0

for all x1 , x2 , y1 , y2 in [0, 1] such that x1 ≤ x2 and y1 ≤ y2 .

Let Π denote the copula for independent random variables, i. e., Π(x, y) = xy for
all (x, y) ∈ [0, 1]2 . For any copula C we have W (x, y) = max(0, x+y−1) ≤ C(x, y) ≤
min(x, y) = M (x, y) for all (x, y) in [0, 1]2 , where M and W are themselves copulas.
For a complete survey on copulas, see [22].
Given a copula C, let C (respectively, Ĉ) denote the survival function (respec-
tively, survival copula) associated with C, i. e., C(x, y) = 1 − x − y + C(x, y) (respec-
tively, Ĉ(x, y) = C(1 − x, 1 − y)). Whereas Ĉ is always a copula, C never is since
condition (C1) does not hold.
In the following, given bivariate functions A and B with a common domain K,
let A ≤ B denote the pointwise inequality A(x, y) ≤ B(x, y) for every (x, y) in K
(and similarly for “≥” and “=”).

3. THE TRANSFORMATIONS
3.1. Constructions and examples
Let (X1 , Y1 ) and (X2 , Y2 ) be two independent vectors of uniform (0, 1) random vari-
ables with common copula D. Let X(1) , X(2) and Y(1) , Y(2) be their corresponding
order statistics. Consider the random vector
(
(X(1) , Y(2) ), with probability 1/2
(Z1 , Z2 ) =
(X(2) , Y(1) ), with probability 1/2.

The distribution of (Z1 , Z2 ) is then given by

1 1
H1 (x, y) = P(X(1) ≤ x, Y(2) ≤ y) + P(X(2) ≤ x, Y(1) ≤ y)
2 2
1
= {P(Y(2) ≤ y) − P(X(1) > x, Y(2) ≤ y) + P(X(2) ≤ x)
2
−P(X(2) ≤ x, Y(1) > y)}
1© 2 ª
= y − (y − D(x, y))2 + x2 − (x − D(x, y))2
2
= D(x, y) {x + y − D(x, y)}
= D(x, y) {1 − (1 − x − y + D(x, y))}
© ª
= D(x, y) 1 − D(x, y) .
994 A. DOLATI AND M. ÚBEDA-FLORES

Now, for α ∈ [0, 1], consider the random pair (T1 , T2 ), defined by
(
(Z1 , Z2 ), with probability α
(T1 , T2 ) =
(X1 , Y1 ), with probability 1 − α.

Then the distribution of (T1 , T2 ), denoted by Cα [D], is given by

© ª
Cα [D] (x, y) = αD(x, y) 1 − D(x, y) + (1 − α)D(x, y)
© ª
= D(x, y) 1 − αD(x, y) .

Note that the bivariate distribution function Cα [D] satisfies the boundary conditions
(C1), and hence it is a copula.
If we consider the random vector
(
(X(1) , Y(1) ), with probability 1/2
(Z1 , Z2 ) =
(X(2) , Y(2) ), with probability 1/2,

then the distribution function of (Z1 , Z2 ) is given by

1 1
H2 (x, y) = P(X(1) ≤ x, Y(1) ≤ y) + P(X(2) ≤ x, Y(2) ≤ y)
2 2
1© ª
= 1 − P(X(1) > x) − P(Y(1) > y) + P(X(1) > x, Y(1) > y) + D2 (x, y)
2
1© ª
= 1 − (1 − x)2 − (1 − y)2 + (1 − x − y + D(x, y))2 + D2 (x, y)
2
= Π(x, y) + D(x, y)D(x, y).

The distribution function of the pair

(
(Z1 , Z2 ), with probability α
(T1 , T2 ) =
(X1 , X2 ), with probability 1 − α,

denoted by Cα∗ [D] – and which is copula as well – is then given by

© ª
Cα∗ [D] (x, y) = α Π(x, y) + D(x, y)D(x, y) + (1 − α)Π(x, y)
= Π(x, y) + αD(x, y)D(x, y).

In short, we have proved the following result.

Theorem 3.1. For any copula D, the functions defined from [0, 1]2 onto [0, 1] by

Cα [D] (x, y) = D(x, y)[1 − αD(x, y)], (1)

and
Cα∗ [D] (x, y) = Π(x, y) + αD(x, y)D(x, y), (2)
where 0 ≤ α ≤ 1, are copulas.
Copulas Based on Pairs of Order Statistics 995

Each transformation (1) and (2) is “unique”, in the sense that given a copula D,
this generates a unique copula – or a family of copulas – (see examples below) under
(1) and (2). We prove it in the following result.

Theorem 3.2. Let D1 and D2 be two copulas such that Cα [D1 ] = Cα [D2 ] (respec-
tively, Cα∗ [D1 ] = Cα∗ [D2 ]) for every α ∈ [0, 1]. Then D1 = D2 .

P r o o f . Let D1 and D2 be two copulas such that Cα [D1 ] = Cα [D2 ] (in the case
Cα∗ [D1 ] = Cα∗ [D2 ] the result can be proved similarly). Suppose α 6= 0 (the case α = 0
is trivial). Then we have D1 (1 − αD1 ) = D2 (1 − αD2 ), which is equivalent to

[D2 (x, y) − D1 (x, y)] {1 − α(1 − x − y) − α[D2 (x, y) + D1 (x, y)]} = 0 (3)

for every (x, y) in [0, 1]2 . Suppose there exists a point (x0 , y0 ) in [0, 1]2 such that
– without loss of generality – D1 (x0 , y0 ) < D2 (x0 , y0 ). Then the equality in (3) is
equivalent to 1 − α(1 − x0 − y0 ) − α[D2 (x0 , y0 ) + D1 (x0 , y0 )] = 0, i. e., D2 (x0 , y0 ) +
D1 (x0 , y0 ) = x0 + y0 − 1 + 1/α. Since 1 ≤ 1/α and D1 (x0 , y0 ) < D2 (x0 , y0 ) ≤ y0 ,
we have the following chain of inequalities: D2 (x0 , y0 ) ≥ x0 + y0 − D1 (x0 , y0 ) >
x0 + y0 − y0 = x0 . This is absurd, and we conclude that D1 = D2 . ¤

We now provide several examples.

Example 3.3. Consider the copula M . Since M (x, y) = M (1 − x, 1 − y) for every

(x, y) in [0, 1]2 , we have M (x, y)M (x, y) = M (x, y) − Π(x, y). The copulas generated
by (1) and (2) are then given by

Cα [M ] (x, y) = αΠ(x, y) + (1 − α)M (x, y), (4)

and

Cα∗ [M ] (x, y) = αM (x, y) + (1 − α)Π(x, y), (5)

respectively. Copulas of the form (4) and (5) belong to the known Fréchet–Mardia
family of copulas (see [22] for more details).

Example 3.4. Consider the copula W . Since W (x, y)W (x, y) = 0 for every (x, y)
in [0, 1]2 , for each α ∈ [0, 1] we have

Cα [W ] (x, y) = W (x, y),

and

Cα∗ [W ] (x, y) = Π(x, y).

Example 3.5. Consider the product copula Π. Then we have

Cα [Π] (x, y) = xy[1 − α(1 − x)(1 − y)], (6)

996 A. DOLATI AND M. ÚBEDA-FLORES

and

Cα∗ [Π] (x, y) = xy[1 + α(1 − x)(1 − y)]. (7)

Copulas given by (6) and (7) constitute the known Farlie–Gumbel–Morgenstern
(FGM, for short) family of copulas, which is usually written as

E(x, y) = xy[1 + α(1 − x)(1 − y)], (8)

with α ∈ [−1, 1] (see [9] and the references therein).

Remark 3.6. Let J be the family of copulas given by J(x, y) = xy + αf (x)g(y),

with f and g two real functions and α ∈ R\{0}. Conditions under f , g and α in order
that J is a copula are given in [23]. If we compare J with Cα∗ [D], i. e., J = Cα∗ [D], for
some copula D and appropriate α, we obtain D(x, y)[1 − x − y + D(x, y)] = f (x)g(y),
whose solution for D is
p
x + y − 1 + (1 − x − y)2 + 4f (x)g(y)
D(x, y) = .
2
For example, taking f (t) = g(t) = t(1 − t) for all t in [0, 1], we obtain D(x, y) = xy
– recall Example 3.5.

3.2. Some properties

In what follows, we provide several properties on symmetry, invariance, ordering,
and measures of association of the copulas given by (1) and (2).
A copula D is Cα -invariant (respectively, Cα∗ -invariant) if Cα [D] = D (respec-
tively, Cα∗ [D] = D). The following result shows for which copulas D the transfor-
mations (1) and (2) are Cα -invariant and Cα∗ -invariant, respectively.

Theorem 3.7. Given a copula D, for each α ∈ [0, 1]:

(i) D is Cα -invariant if, and only if, D = W .

(ii) D is Cα∗ -invariant if, and only if, D is the Plackett family of copulas, which is
given by
p
1 − α(1 − x − y) − (α(1 − x − y) − 1)2 − 4αxy
D(x, y) = (9)
2α
for α > 0, and D = Π if α = 0.

P r o o f . By definition of Cα [D], we have Cα [D] = D if, and only if, D(x, y)(1 − x −
y + D(x, y)) = 0, or equivalently, D(x, y) = 0 or D(x, y) = x + y − 1, i. e., D = W ,
which proves part (i).
To prove part (ii), note that if α > 0 (the case α = 0 is trivial and we omit
it) Cα∗ [D] = D is equivalent to the following equality: αD2 (x, y) − (1 − α(1 − x −
Copulas Based on Pairs of Order Statistics 997

y))D(x, y) + xy = 0, whose solution for D is given by (9) – see [22] for more details
on this family of copulas. This completes the proof. ¤

Observe that if α = 1 in the copula D given by (9), we obtain

p
x + y − (x − y)2 x + y − |x − y|
D(x, y) = = ,
2 2
i. e., D = M .
For each α ∈ [0, 1], we have Cα∗ [D] = D if, and only if, Cα [D] = Π (and Cα [D] = D
if, and only if, Cα∗ [D] = Π). This fact and Theorem 3.7 lead us to the following

Corollary 3.8. Given a copula D, for each α ∈ [0, 1] we have:

(i) Cα [D] = Π if, and only if, D is the family of copulas given by (9).
(ii) Cα∗ [D] = Π if, and only if, D = W .

If the copula D is symmetric, i. e., D(x, y) = D(y, x) for all x, y ∈ [0, 1], then (1)
and (2) are symmetric as well. A copula C is radially symmetric if C = Ĉ [21, 22].
The following result shows that the transformations (1) and (2) preserve the radially
symmetry property of a given copula.

Theorem 3.9. For every radially symmetric copula D, the transformations Cα [D]
and Cα∗ [D] are radially symmetric for each α ∈ [0, 1].

P r o o f . By definition, we have

Ĉα [D] (x, y) = x + y − 1 + Cα [D] (1 − x, 1 − y)

© ª
= x + y − 1 + D(1 − x, 1 − y) 1 − αD(1 − x, 1 − y)
= x + y − 1 + D(1 − x, 1 − y) − αD(1 − x, 1 − y)D(1 − x, 1 − y)
n o
= D̂(x, y) 1 − αD̂(x, y) ,

where the last line follows from the fact that D(1 − x, 1 − y) = P(X ≤ 1 − x, Y ≤
1 − y) = P(1 − X > x, 1 − Y > y) = D̂(x, y) and D(1 − x, 1 − y) = D̂(x, y). Thus
Ĉα [D] = Cα [D̂] = Cα [D], which completes the proof. ¤

If C1 and C2 are two copulas, we say that C2 is more concordant than C1 (written
C1 ≺c C2 ) if C1 ≤ C2 . A copula C is positively quadrant dependent (written PQD)
if Π ≺c C, and negatively quadrant dependence (NQD) if C ≺c Π. A totally ordered
parametric family {Cα } of copulas is positively ordered if Cα1 ≺c Cα2 whenever
α1 ≤ α2 ; and negatively ordered if Cα2 ≺c Cα1 whenever α1 ≤ α2 [15, 22]. For the
families (1) and (2) of copulas we have the following results – the proof of Theorem
3.10 is simple, and we omit it.
998 A. DOLATI AND M. ÚBEDA-FLORES

Theorem 3.10. Given a copula D, we have:

(i) The parametric family {Cα [D]} (respectively, {Cα∗ [D]}) of copulas is negatively
(respectively, positively) ordered.
(ii) Cα [D] ≺c D and Π ≺c Cα∗ [D] for all α ∈ [0, 1], i. e., Cα∗ [D] is PQD for every
α ∈ [0, 1].
Theorem 3.11. Let D1 and D2 be two copulas such that D1 ≺c D2 . Then we have
Cα [D1 ] ≺c Cα [D2 ] and Cα∗ [D1 ] ≺c Cα∗ [D2 ] for every α ∈ [0, 1].

P r o o f . We first prove the result for the transformation (1). Note that, for a given
copula D, we have Cα [D] = αD(1 − D) + (1 − α)D. So we only need to show
that D1 ≺c D2 implies that D1 (1 − D1 ) ≤ D2 (1 − D2 ); but this last inequality is
equivalent to
D1 (x, y) [x + y − D1 (x, y)] − D2 (x, y) [x + y − D2 (x, y)] ≤ 0,
that is,
[D2 (x, y) − D1 (x, y)] [x − D1 (x, y) + y − D2 (x, y)] ≥ 0.
Since D1 (x, y) ≤ x and D2 (x, y) ≤ y, the result follows.
On the other hand, since D1 ≺c D2 implies that D1 ≺c D2 , and thus D1 D1 ≤
D2 D2 , it follows that Cα∗ [D1 ] ≺c Cα∗ [D2 ], which completes the proof. ¤

We note that, for a given copula D, the transformation Cα [D] may be PQD or
NQD. For instance, from part (ii) of Theorem 3.10, if a copula D is NQD, then Cα [D]
is NQD. And, for a given copula D, if B ≺c D, where B is a copula given by (9),
using Theorem 3.11 and part (i) of Corollary 3.8, we have that Π = Cα [B] ≺c Cα [D],
i. e., Cα [D] is PQD.
As a consequence of our results, transformations of type (1) (respectively, (2))
decrease (respectively, increase) the degree of dependence of the copula D. A useful
family for modeling could be a convex linear combination of Cα [D] and Cα∗ [D] –
which is a copula – namely
C(x, y) = βCα [D](x, y) + (1 − β)Cα∗ [D](x, y) for all (x, y) ∈ [0, 1]2 , β ∈ [0, 1]. (10)
Observe that, for instance, taking D = Π in (10), we easily obtain C(x, y) = xy[1 +
α(1 − 2β)(1 − x)(1 − y)], i. e., the FGM family of copulas; which is PQD if, and only
if, β ≤ 1/2, and NQD when β ≥ 1/2.
The population version of three of the most common nonparametric measures
of association between the components of a continuous random pair (X, Y ) are
Kendall’s tau (τ ), Spearman’s rho (ρ), and Gini’s gamma (γ). Such measures are
called measures of concordance since they satisfy a set of axioms due to Scarsini [24].
The coefficients ρ and γ are based on average quadrant dependence and τ on expected
quadrant dependence [4], and depend only on the copula C of the pair (X, Y ), and
are given by Z Z 1 1
τ (C) = 4 C(x, y) dC(x, y) − 1, (11)
0 0
Copulas Based on Pairs of Order Statistics 999

Z 1Z 1
ρ(C) = 12 [C(x, y) − xy] dxdy, (12)
0 0

and Z 1Z 1
γ(C) = 8 [C(x, y) − xy] dA(x, y), (13)
0 0

where A denotes the copula (M + W )/2 (for a complete study on measures of

association, see [22] and the references therein). The direct computation of these
measures for the copulas in (1) and (2) does not provide too much information;
however, we have the following result, in which we find relationships among these
measures for the copulas Cα and Cα∗ .

Theorem 3.12. Let τ be the measure given by (11), and let κ denote the set of the
measures given by (12) and (13), i. e., κ ∈ {ρ, γ}. Then, for a given copula D and
for each α ∈ [0, 1], we have:

(i) −1 ≤ τ (Cα [D]) ≤ [(1 − α)(3 − α)]/3 and −1 ≤ κ(Cα [D]) ≤ 1 − α;

(ii) 0 ≤ τ (Cα∗ [D]) ≤ [α(α + 2)]/3 and 0 ≤ κ(Cα∗ [D]) ≤ α;

(iii) κ(Cα [D]) + κ(Cα∗ [D]) = κ(D).

P r o o f . First we recall that given two copulas C1 and C2 such that C1 ≺c C2 , we

have τ (C1 ) ≤ τ (C2 ), ρ(C1 ) ≤ τ (C2 ), and γ(C1 ) ≤ γ(C2 ) [22].
Since any copula D satisfies that W ≺c D ≺c M , from Theorem 3.11 and Exam-
ples 3.3 and 3.4, we have W ≺c Cα [D] ≺c (αΠ + (1 − α)M ) and Π ≺c Cα∗ [D] ≺c
(αM + (1 − α)Π). Thus, the lower bound for the transformation (1) for the three
measures is −1, and easy computations give us the upper bounds, which proves part
(i). A similar argument permits to prove part (ii). Finally, part (iii) follows from
the fact that (Cα [D] − Π) + (Cα∗ [D] − Π) = (D − Π), then apply it to the expressions
in (12) and (13). This completes the proof. ¤

We finish this section providing a procedure for generating random value (T1 , T2 ),
from copulas of types (1) and (2). For the first type, consider the following algorithm:

Step 1. Generate a uniform (0, 1) random variable S1 ;

Step 2. generate independent random pairs (X1 , Y1 ) and (X2 , Y2 ), from the given
copula D and sort them into X(1) ≤ X(2) and Y(1) ≤ Y(2) .
Step 3. If S1 > α, take (T1 , T2 ) = (X1 , Y1 );
Step 4. else, generate a uniform (0, 1) random variable S2 and take (T1 , T2 ) =
(X(1) , Y(2) ) if S2 < 1/2, otherwise take (T1 , T2 ) = (X(2) , Y(1) ).

A similar algorithm can be used for the second transformation.

1000 A. DOLATI AND M. ÚBEDA-FLORES

3.3. Additional transformations

A known limitation of the FGM family of copulas given by (8) is that the range
of Kendall’s tau and Spearman’s rho are limited to [−2/9, 2/9] and [−1/3, 1/3],
respectively; so that it does not allow for modeling of strong dependence [15]. Ap-
plying the transformations (1) and (2) to the FGM family of copulas Dθ (x, y) =
xy[1 + θ(1 − x)(1 − y)] with θ ∈ [−1, 1], we obtain two new iterated FGM distribu-
tions of the form

Cα [Dθ ] (x, y) = xy[1 + θ(1 − x)(1 − y)][1 − α(1 − x)(1 − y)(1 + θxy)], (14)

and

Cα∗ [Dθ ] (x, y) = xy+αxy[1+θ(1−x)(1−y)]{(1−x−y+xy[1+θ(1−x)(1−y)]}, (15)

respectively, where θ ∈ [−1, 1] and α ∈ [0, 1] – see [9] for other iterated FGM
Distributions. After some algebra, we obtain
µ ¶
θ 1 θ θ2
ρ(Cα [Dθ ]) = − + + α,
3 3 6 75
µ ¶
1 θ θ2
ρ(Cα∗ [Dθ ]) = + + α,
3 6 75
µ ¶ µ 2 ¶
2θ 4θ2 θ 1 θ θ 2
τ (Cα [Dθ ]) = + + + α θ−
2
+ + α,
9 11025 450 225 75 9 9
µ ¶ µ ¶
4θ2 θ 1 2θ2 θ 2
τ (Cα∗ [Dθ ]) = + + α2 θ + + + α.
11025 450 225 225 9 9

We have the following ranges: ρ(Cα [Dθ ]) ∈ [−0.51b 3, 0.b

3], ρ(Cα∗ [Dθ ]) ∈ [0, 0.51b
3],
b ∗
τ (Cα [Dθ ]) ∈ [−0.3492517, 0.2], and τ (Cα [Dθ ]) ∈ [0, 0.3492517]. So that, for in-
stance, using (14), Spearman’s rho can be decreased down to −0.51b 3 at (α, θ) =
(1, −1); and using (15), Spearman’s rho can be increased up to 0.51b 3 at α = θ = 1.

4. DISCUSSION
In this paper, we have introduced and studied two new transformations of copulas.
Other transformations of copulas into copulas can be defined. For example, for a
given copula D, we define the transformations

D(x, y)
Hβ (x, y)[D] = (16)
1 − βD(x, y)

and
Nδ (x, y)[D] = D(x, y) exp[δD(x, y)] (17)
for every (x, y) in [0, 1] with β, δ ∈ R. Observe that
2

xy
Hβ (x, y)[Π] = ,
1 − β(1 − x)(1 − y)
Copulas Based on Pairs of Order Statistics 1001

with β ∈ [−1, 1], is the Ali–Mikhail–Haq family of copulas [1]; and

Nδ (x, y)[Π] = xy exp[δ(1 − x)(1 − y)],

with δ ∈ [−1, 1], is a copula studied in [7]. We also note that Nδ [M ] is not a copula.
Observe that the transformations (1), (2), (16), and (17) provide a (single) copula
or a one-parametric family of copulas. We can also define transformations of copulas
which provide multi-parametric families of copulas. For instance

Pα,β,θ1 ,θ2 (x, y)[D] = D(x, y)[1 + αD(x, y)]θ1 [1 + βD(x, y)]θ2 , (18)

with α, β, θ1 , θ2 real numbers. In particular, we have

Pα,β,θ1 ,θ2 (x, y)[Π] = xy[1 + α(1 − x)(1 − y)]θ1 [1 + β(1 − x)(1 − y)]θ2 ,

with −1 ≤ α, β ≤ 1 and |θ1 | + |θ2 | = 1, is a copula studied in [7].

Conditions on the parameters or on the copula D so that the transformations
(16) – (18) are copulas is the subject of further work.

ACKNOWLEDGEMENT
The authors wish to thank Prof. Christian Genest and two anonymous referees for some
enlightening remarks on an earlier version of this paper. The first author thanks the
research council of Yazd University (Iran) for support. The second author acknowledges
the support of the Ministerio de Educación y Ciencia (Spain) and FEDER, under research
project MTM2006-12218.

(Received July 9, 2008.)

REFERENCES

[1] M. M. Ali, N. N. Mikhail, and M. S. Haq: A class of bivariate distributions including

the bivariate logistic. J. Multivariate Anal. 8 (1978), 405–412.
[2] E. Alvoni, P. L. Papini, and F. Spizzichino: On a class of transformations of copulas
and quasi-copulas. Fuzzy Sets and Systems 50 (2009), 334–343.
[3] R. Baker: An-order-statistics-based method for constructing distributions with fixed
marginals. J. Multivariate Anal. 99 (2008), 2312–2327.
[4] J. Behboodian, A. Dolati, and M. Úbeda-Flores: Measures of association based on
average quadrant dependence. J. Probab. Statist. Sci. 3 (2005), 161–173.
[5] G. Beliakov, A. Pradera, and T. Calvo: Aggregation Functions: A Guide for Practi-
tioners. Springer, New York 2007.
[6] T. Calvo, G. Mayor, and R. Mesiar (eds.): Aggregation Operators: New Trends and
Applications. Physica-Verlag, Heidelberg 2002.
[7] C. M. Cuadras: Constructing copula functions with weighted geometric means. J.
Statist. Plann. Inference 139 (2009), 3766–3772.
[8] B. De Baets, H. De Meyer, and S. Dı́az: On an idempotent transformation of aggre-
gation functions and its application on absolutely continuous Archimedean copulas.
Fuzzy Sets and Systems 160 (2009), 733–751.
1002 A. DOLATI AND M. ÚBEDA-FLORES

[9] D. Drouet Mari and S. Kotz: Correlation and Dependence. Imperial College Press,
London 2001.
[10] F. Durante: Construction of non-exchangeable bivariate distribution functions.
Statist. Papers 50 (2009), 383–391.
[11] F. Durante and C. Sempi: Copula and semicopula transforms. Internat. J. Math.
Math. Sci. 4 (2005), 645–655.
[12] F. Durante, R. Mesiar, P. L. Papini, and C. Sempi: 2-increasing binary aggregation
operators. Inform. Sci. 177 (2007), 111-129.
[13] V. Durrleman, A. Nikeghbali, and T. Roncalli: A Simple Transformation of Copulas.
Technical Report. Groupe de Research Operationnelle Credit–Lyonnais 2000.
[14] P. Hájek and R. Mesiar: On copulas, quasi-copulas and fuzzy logic. Soft Computing
12 (2008), 1239–1243.
[15] H. Joe: Multivariate Models and Dependence Concepts. Chapman and Hall, London
1997.
[16] E. P. Klement, R. Mesiar, and E. Pap: Triangular Norms. Kluwer, Dordrecht 2000.
[17] E. P. Klement, R. Mesiar, and E. Pap: Archimax copulas and invariance under trans-
formations. C.R. Acad. Sci. Paris, Ser. I 340 (2005), 755–758.
[18] E. P. Klement, R. Mesiar, and E. Pap: Transformations of copulas. Kybernetika 41
(2005), 425–434.
[19] R. Mesiar and C. Sempi: Ordinal sums and idempotents of copulas. Aequationes
Math. (to appear)
[20] P. M. Morillas: A method to obtain new copulas from a given one. Metrika 61 (2005),
169–184.
[21] R. B. Nelsen: Some concepts of bivariate symmetry. J. Nonparametric Statist. 3
(1993), 95–101.
[22] R. B. Nelsen: An Introduction to Copulas. Second Edition. Springer, New York 2006.
[23] J. A. Rodrı́guez-Lallena and M. Úbeda-Flores: A new class of bivariate copulas.
Statist. Probab. Lett. 66 (2004), 315–325.
[24] M. Scarsini: On measures of concordance. Stochastica 8 (1984), 201–218.
[25] A. Sklar: Fonctions de répartition à n dimensions et leurs marges. Publ. Inst. Statist.
Univ. Paris 8 (1959), 229–231.
[26] A. Sklar: Random variables, joint distributions, and copulas. Kybernetika 9 (1973),
449–460.

Ali Dolati, Department of Statistics, Faculty of Science, Yazd University, Yazd,

89195-741. Iran.
e-mail: [email protected]
Manuel Úbeda-Flores, Departamento de Estadı́stica y Matemática Aplicada, Univer-
sidad de Almerı́a, Carretera de Sacramento s/n, 04120 La Cañada de San Urbano,
Almerı́a. Spain.
e-mail: [email protected]