4.1 Distribution and Density Function
Let
be a random vector.
The cumulative distribution function (cdf) of
is defined by
For continuous
,
there exists a nonnegative probability density function (pdf)
, such that
![\begin{displaymath}
F(\undertilde x) = \int ^{\undertilde x}_{-\infty }
f (\undertilde u)d{\undertilde u}.
\end{displaymath}](mvahtmlimg1095.gif) |
(4.1) |
Note that
Most of the
integrals appearing below are multidimensional. For instance,
means
Note also that the cdf
is differentiable with
For discrete
, the values of this random variable are concentrated on a
countable or finite set of points
, the
probability of events of the form
can then be computed as
If we partition
as
with
and
, then the function
![\begin{displaymath}F_{X_{1}}(\undertilde x_1)=P(X_1\le \undertilde x_1)
=F(x_{11},\ldots ,x_{1k},\infty ,\ldots, \infty) \end{displaymath}](mvahtmlimg1106.gif) |
(4.2) |
is called the marginal cdf.
is called the joint cdf.
For continuous
the marginal pdf can be computed from the joint density
by ``integrating out'' the variable not of interest.
![\begin{displaymath}f_{X_{1}}(\undertilde x_1) = \int ^\infty _{-\infty }f
(\undertilde x_1,\undertilde x_2) d\undertilde x_2. \end{displaymath}](mvahtmlimg1108.gif) |
(4.3) |
The conditional pdf of
given
is given as
![\begin{displaymath}f(\undertilde x_2\mid \undertilde x_1) = \frac{f(\undertilde
x_1,\undertilde x_2) }{f_{X_{1}}(\undertilde x_1)}\cdotp \end{displaymath}](mvahtmlimg1110.gif) |
(4.4) |
EXAMPLE 4.1
Consider the pdf
![$f(x_1,x_2)$](mvahtmlimg1112.gif)
is a density since
The marginal densities
are
The conditional densities
are therefore
Note that these conditional pdf's are nonlinear in
![$x_{1}$](mvahtmlimg1116.gif)
and
![$x_{2}$](mvahtmlimg1117.gif)
although the joint pdf has a simple (linear) structure.
Independence of two random variables is defined as follows.
DEFINITION 4.1
![$X_1$](mvahtmlimg14.gif)
and
![$X_2$](mvahtmlimg13.gif)
are independent
iff
![$f(x) = f(x_1,x_2) = f_{X_{1}}(x_1) f_{X_{2}}(x_2)$](mvahtmlimg1118.gif)
.
That is,
and
are independent if
the conditional pdf's are equal to the marginal
densities, i.e.,
and
. Independence can be interpreted
as follows: knowing
does not change the probability assessments
on
, and conversely.
1mm
Different joint pdf's
may have the same marginal pdf's.
EXAMPLE 4.2
Consider the pdf's
and
We compute in both cases the marginal pdf's as
Indeed
Hence we obtain identical marginals from
different joint distributions!
Let us study the concept of independence using the bank notes example.
Consider the variables
(lower inner frame) and
(upper
inner frame).
From Chapter 3, we already know that they
have significant correlation, so they are almost surely not independent.
Kernel estimates of the marginal densities,
and
, are given in Figure 4.1.
In Figure 4.2 (left) we show the product of these two
densities.
The kernel density technique was presented in Section 1.3.
If
and
are independent, this product
should be roughly equal to
, the estimate of the joint density of
.
Comparing the two graphs in Figure 4.2
reveals that the two densities are different.
The two variables
and
are therefore not independent.
Figure 4.1:
Univariate estimates of the density of
(left) and
(right) of the bank notes.
MVAdenbank2.xpl
|
Figure 4.2:
The product of univariate density estimates
(left) and the joint density estimate (right) for
(left) and
of the bank notes.
MVAdenbank3.xpl
|
An elegant concept of connecting marginals with joint cdfs is given by
copulas.
Copulas are important in Value-at-Risk calculations and
are an essential tool in quantitative finance (Härdle et al.; 2002).
For simplicity of presentation we concentrate on the
dimensional case.
A 2-dimensional copula is a function
with the following properties:
The usage of the name ``copula'' for the function
is explained by the
following theorem.
THEOREM 4.1 (Sklar's theorem)
Let
![$F$](mvahtmlimg64.gif)
be a joint distribution function with marginal distribution
functions
![$F_{X_1}$](mvahtmlimg1143.gif)
and
![$F_{X_2}$](mvahtmlimg1144.gif)
. Then there exists a copula
![$C$](mvahtmlimg1142.gif)
with
![\begin{displaymath}
F(x_1,x_2) = C\{ F_{X_1}(x_1),F_{X_2}(x_2)\}
\end{displaymath}](mvahtmlimg1145.gif) |
(4.5) |
for every
![$x_1,x_2 \in \mathbb{R}$](mvahtmlimg1146.gif)
. If
![$F_{X_1}$](mvahtmlimg1143.gif)
and
![$F_{X_2}$](mvahtmlimg1144.gif)
are
continuous, then
![$C$](mvahtmlimg1142.gif)
is unique.
On the other hand, if
![$C$](mvahtmlimg1142.gif)
is a copula and
![$F_{X_1}$](mvahtmlimg1143.gif)
and
![$F_{X_2}$](mvahtmlimg1144.gif)
are distribution functions, then
the function
![$F$](mvahtmlimg64.gif)
defined by (
4.5) is a joint distribution function
with marginals
![$F_{X_1}$](mvahtmlimg1143.gif)
and
![$F_{X_2}$](mvahtmlimg1144.gif)
.
With Sklar's Theorem, the use of the name ``copula''
becomes obvious.
It was chosen to describe ``a function that
links a multidimensional distribution to its one-dimensional
margins'' and appeared in the mathematical literature for the first
time in Sklar (1959).
EXAMPLE 4.3
The
structure of independence implies that
the product of the distribution functions
![$F_{X_1}$](mvahtmlimg1143.gif)
and
![$F_{X_2}$](mvahtmlimg1144.gif)
equals their joint distribution function
![$F$](mvahtmlimg64.gif)
,
![\begin{displaymath}
F(x_1,x_2) = F_{X_1}(x_1) \cdot F_{X_2}(x_2).
\end{displaymath}](mvahtmlimg1147.gif) |
(4.6) |
Thus, we obtain the
independence copula
![$C = \Pi$](mvahtmlimg1148.gif)
from
THEOREM 4.2
Let
![$X_1$](mvahtmlimg14.gif)
and
![$X_2$](mvahtmlimg13.gif)
be random variables with continuous
distribution functions
![$F_{X_1}$](mvahtmlimg1143.gif)
and
![$F_{X_2}$](mvahtmlimg1144.gif)
and the joint distribution
function
![$F$](mvahtmlimg64.gif)
.
Then
![$X_1$](mvahtmlimg14.gif)
and
![$X_2$](mvahtmlimg13.gif)
are independent if and only if
![$C_{X_1, X_2} = \Pi$](mvahtmlimg1150.gif)
.
PROOF:
From Sklar's Theorem we know that there exists an
unique copula
with
![\begin{displaymath}
P (X_1 \le x_1, X_2 \le x_2) = F(x_1,x_2) =
C\{F_{X_1}(x_1),F_{X_2}(x_2)\} \, .
\end{displaymath}](mvahtmlimg1151.gif) |
(4.7) |
Independence can be seen using (4.5) for the joint
distribution function
and the definition of
,
![\begin{displaymath}
F(x_1,x_2) = C\{F_{X_1}(x_1),F_{X_2}(x_2)\} = F_{X_1}(x_1) F_{X_2}(x_2) \; .
\end{displaymath}](mvahtmlimg1153.gif) |
(4.8) |
EXAMPLE 4.4
The Gumbel-Hougaard family of
copulas (Nelsen; 1999) is given by the function
![\begin{displaymath}
C_{\theta}(u, v) = \exp \left\{ - \left[ (-\ln u)^{\theta}
+ (-\ln v)^{\theta} \right]^{1 / \theta} \right\} \; .
\end{displaymath}](mvahtmlimg1154.gif) |
(4.9) |
The parameter
![$\theta$](mvahtmlimg581.gif)
may take all values in the interval
![$[1,\infty)$](mvahtmlimg1155.gif)
.
The Gumbel-Hougaard copulas are suited to describe bivariate extreme
value distributions.
For
, the expression (4.9)
reduces to the product copula, i.e.,
.
For
one finds for the Gumbel-Hougaard copula:
where the function
![$M$](mvahtmlimg62.gif)
is also a copula such that
![$C(u,v) \le M(u,v)$](mvahtmlimg1160.gif)
for
arbitrary copula
![$C$](mvahtmlimg1142.gif)
. The copula
![$M$](mvahtmlimg62.gif)
is called the
Fréchet-Hoeffding upper bound.
Similarly, we obtain the Fréchet-Hoeffding lower bound
which satisfies
for any other copula
.
Summary
![$\ast$](mvahtmlimg108.gif)
-
The cumulative distribution function (cdf) is defined as
.
![$\ast$](mvahtmlimg108.gif)
-
If a probability density function (pdf)
exists then
.
![$\ast$](mvahtmlimg108.gif)
-
The pdf integrates to one, i.e.,
.
![$\ast$](mvahtmlimg108.gif)
-
Let
be partitioned into sub-vectors
and
with joint cdf
. Then
is the marginal cdf of
.
The marginal pdf of
is obtained by
. Different joint pdf's
may have the same marginal pdf's.
![$\ast$](mvahtmlimg108.gif)
-
The conditional pdf of
given
is defined as
![$\ast$](mvahtmlimg108.gif)
-
Two random variables
and
are called independent iff
. This is equivalent to
.
![$\ast$](mvahtmlimg108.gif)
- Different joint pdf's may have identical marginal pdf's.