Civil Engineering Reference
In-Depth Information
3.2
The Time-Varying Copula for Dependence Parameters
We commonly use Pearson's correlation coefficient
ρ
t
to describe the dependence
structure in the dynamic Gaussian copula and dynamic Student-
t
copula. On the
other hand, we often use the Kendall's
τ
t
on the Gumbel and Clayton copulas. We
assumed that the dependence parameters rely on past dependence and historical
information
(
m
t
−
1
−
0
.
5
)(
w
t
−
1
−
0
.
5
)
. The dependence process of the Gaussian and
Student-
t
are, therefore,
ρ
t
=
Λ
(
α
c
+
β
c
ρ
t
−
1
+
γ
c
(
m
t
−
1
−
0
.
5
)(
w
t
−
1
−
0
.
5
))
.
(11)
The dependence process of the Gumbel is
τ
t
=
Λ
(
α
c
+
β
c
τ
t
−
1
+
γ
c
(
m
t
−
1
−
0
.
5
)(
w
t
−
1
−
0
.
5
))
.
(12)
In the conditional dependence, we assumed that
ρ
t
and
τ
t
determined from its past
level,
ρ
(
t
−
1
)
and
τ
(
t
−
1
)
. The parameter of
β
c
captures the persistent effect and
γ
c
can captures historical information.
, denotes the logistic
transformation, which is used to ensure the dependence parameters fall within the
interval
Λ
=
−
ln
[(
1
−
x
t
)
/
(
1
+
x
t
)]
.
While we change the historical information
(
−
1
,
1
)
(
m
0
,
t
−
1
−
0
.
5
)(
w
e
,
t
−
1
−
0
.
5
)
to
1
10
10
i
, captures the variability of the dependence.
We proposed time-varying dependence processes for Clayton copula as
|
m
t
−
1
−
w
t
−
1
|
∑
=
1
10
i
=
1
|
w
t
−
1
−
m
t
−
1
|
1
10
τ
t
=
Π
α
c
+
β
c
τ
t
−
1
+
γ
c
(13)
e
−
x
)
−
1
denotes the logistic transformation.
Π
=(
1
+
3.3
The Estimation and Calibration of the Copula
IFM method is used to estimate copula-based GARCH mode and get the parameters.
The method function is
T
t
=
1
ln
f
it
(
k
i
,
t
,
θ
it
)
ˆ
θ
it
=
arg max
(14)
T
t
=
1
ln
C
it
(
F
it
(
k
1
,
t
)
,...,
F
nt
(
k
n
,
t
)
,
θ
ct
,
ˆ
ˆ
θ
ct
=
arg max
θ
it
)
.
(15)
Search WWH ::
Custom Search