Information Technology Reference
In-Depth Information
In what follows, we denote by
X
i
,
i
=
1
,...,d
, the coordinate projection of
(X
1
,...,X
d
)
∈ R
d
. Coordinate projections of
=
the multidimensional process
X
Lévy processes are again Lévy processes.
(X
1
,...,X
d
)
=
Proposition 14.1.2
Let X
be a Lévy process with state space
,ν,γ)
.
Assume
|
d
R
Q
>
1
|
z
|
ν(
d
z) <
∞
and characteristic triplet (
.
Then
,
the
z
|
marginal processes X
j
,
j
=
1
,...,d
,
are again Lévy processes with the charac-
teristic triplet (
Q
jj
,ν
j
,γ
j
) where the marginal Lévy measures are defined as
ν
j
(B)
:=
ν(
{
x
∈ R
d
:
x
j
∈
B
\{
0
}}
),
∀
B
∈
B
(
R
),
j
=
1
,...,d.
If all the marginal Lévy processes
X
i
satisfy the martingale condition as
in Lemma 10.1.5,i.e.
e
X
i
are martingales with respect to the filtration of
X
i
,
i
=
1
,...,d
, they are also martingales with respect to the filtration
F
of
X
=
(X
1
,...,X
d
)
.
d
Lemma 14.1.3
Let X be a Lévy process with state space
R
and characteris-
,ν,γ)
.
Assume
,
and
|
z
|
>
1
e
z
j
ν
j
(
d
z) <
tic triplet (
Q
|
z
|
>
1
|
z
|
ν(
d
z) <
∞
∞
,
j
=
1
,...,d
.
Then
,
e
X
j
,
j
=
1
,...,d
,
are martingales with respect to the filtration
F
of
X if and only if
Q
jj
2
e
z
j
z
j
ν
j
(
d
z)
+
γ
j
+
−
1
−
=
0
,j
=
1
,...,d.
R
Proof
As in the proof of Lemma 10.1.5,wehave
E
e
X
s
|
F
t
=
e
X
t
e
(t
−
s)ψ(
−
ie
j
)
.
Therefore, setting
ψ(
−
ie
j
)
=
0,
j
=
1
,...,d
, the Lévy-Khinchine formula (
14.1
)
yields
e
z
j
−
z
j
ν(
d
z)
=
Q
jj
2
+
γ
j
+
−
1
0
,j
=
1
,...,d.
R
d
The result follows with the definition of the marginal Lévy density
ν
j
given in
Proposition
14.1.2
.
14.2 Lévy Copulas
d
We first
gi
ve so
m
e definitions. The
F
-
volume
of
(a, b
]
,
a,b
∈ R
for a function
d
F
:
S
→ R
,
S
⊂ R
is defined by
1
)
N(u)
F(u),
V
F
(
(a, b
]
:=
(
−
)
u
∈{
a
1
,b
1
}×···×{
a
d
,b
d
}
where
N(u)
= |{
k
:
u
k
=
a
k
}|
.
Search WWH ::
Custom Search