com/abstract=1054721
ISSN 12830623
L
EONARD DE VINCI
Estimating the Wishart Ane Stochastic
Correlation Model using the Empirical
Characteristic Function
Jose Da Fonseca Martino Grasselli Florian Ielpo
112007
ESILV, DERMIF, No RR35
Electronic copy available at: http://ssrn.com/abstract=1054721
Estimating the Wishart Ane Stochastic
Correlation Model using the Empirical
Characteristic Function
Jose Da Fonseca
Martino Grasselli
Florian Ielpo
t
dZ
t
_
, (1)
where is the vector of returns and Z
t
R
n
is a vector Brownian motion. Fol
lowing Gourieroux and Sufana (2004), we assume that the quadratic variation of
the risky assets is a matrix
t
which is assumed to satisfy the following dynamics:
d
t
=
_
+M
t
+
t
M
_
dt +
_
t
dW
t
Q+Q
(dW
t
)
t
, (2)
with , M, Q M
n
, invertible, and W
t
M
n
a matrix Brownian motion (
denotes transposition). In the present framework we assume that the above dy
namics are inferred from observed asset price time series, hence the stochastic
dierential equation is written under the historical measure.
Equation (2) characterizes the Wishart process introduced by Bru (1991), and
represents the matrix analogue of the square root meanreverting process. In
order to ensure the strict positivity and the typical meanreverting feature of the
volatility, the matrix M is assumed to be negative semidenite, while satises
= Q
Q, > n 1, (3)
with the real parameter > n 1 (see Bru, 1991 p. 747).
In full analogy with the squareroot process, the term
is related to the
expected longterm variancecovariance matrix
= M
. (4)
4
Moreover, Q is the volatility of the volatility matrix, and its parameters will be
crucial in order to explain some stylized observed eects in equity markets.
Last but not least, Da Fonseca et al. (2006) proposed a very special yet tractable
correlation structure that is able to accommodate the leverage eects found in
nancial time series and option prices. Since it is well known that it is possible
to approximately reproduce observed negative skewness within the Heston (1993)
model by allowing for negative correlation between the noise driving returns and
the noise driving variance, they proposed the following assumption:
Assumption 2.2. The Brownian motions of the asset returns and those driving
the covariance matrix are linearly correlated.
Da Fonseca et al. (2006) proved that Assumption 2.2. leads to the following
relation:
dZ
t
= dW
t
+
_
1
dB
t
, (5)
with Z
t
= (dZ
1
, dZ
2
, . . . , dZ
n
)
.
With this specication, the model is able to generate negative skewness, given
the possibly negative correlation between the noise driving the log returns of the
assets and the matrixsized noise perturbating the covariance matrix. This is
easy to show in the special case of two assets (n = 2), fow which the variance
covariance matrix is given by
t
=
_
11
t
12
t
12
t
22
t
_
. (6)
The correlations between assets returns and their volatilities admit a closed form
expression, highlighting the impact of the parameters on its value and positivity:
corr
_
d log S
1
, d
11
_
=
1
Q
11
+
2
Q
21
_
Q
2
11
+Q
2
21
(7)
corr
_
d log S
2
, d
22
_
=
1
Q
12
+
2
Q
22
_
Q
2
12
+Q
2
22
, (8)
where we recall that
11
(resp.
22
) represents the volatility of the rst
(resp. second) asset. Therefore, the sign and magnitude of the skew eects are
determined by both the matrix Q and the vector . When Q is diagonal, we
obtain the following skews for asset 1 and 2:
corr
_
d log S
1
, d
11
_
=
1
corr
_
d log S
2
, d
22
_
=
2
, (9)
thus allowing a negative skewness for each asset whenever
i
< 0, i (see Gourier
oux and Jasiak (2001) on this point). This correlation structure is similar the
5
one obtained in an Heston model.
Other less general specications close to the WASC model, actually nested within
the WASC correlation structure, have been proposed in the literature. First,
Gourieroux and Sufana (2004) imposed = 0
n
R
n
, a choice that leads to a
zero correlation case (see equations (5), (7) and (8)) by analogy with the well
known properties of the Heston model. With this specication, the log returns
univariate distribution is symmetric.
Second, Buraschi et al. (2006) proposed to impose = (1, 0)
. Their model is
thus able to display negative skewness for asset 1 (resp. asset 2), depending on
the positivity of Q
11
(resp. Q
12
). This model is actually close to the WASC and is
able to display similar features. Their choice of is less restrictive than it seems,
in so far as this parameter is only dened up to a rotation. Thus, their hypothesis
is reduced to [[[[ = 1. With these settings, the vectorsized noise in the returns is
fully generated by the Brownian motions of the covariances W. This hypothesis
having no a priori justication, the WASC model of Da Fonseca et al. (2006)
eliminates it, assuming the more general correlation structure compatible with
an exponential ane characteristic function (see Proposition 1 in Da Fonseca
et al. (2006) on this point).
2.2 The Characteristic functions
In the WASC model, the characteristic function of
t
and Y
t
= log S
t
is an expo
nential ane function of the state variables. For the log returns, the characteristic
function of Y
t+
conditional on Y
t
and
t
is denoted:
Y
t
,
t
(, ) = E
_
e
i,Y
t+
[
t
, Y
t
, (10)
where E[.[
t
, Y
t
] denotes the conditional expected value with respect to the his
torical measure, R
n
, i
2
= 1 and ., .) is the scalar product in R
n
.
Proposition 2.1. (Da Fonseca et al., 2006) The characteristic function of the
asset returns in the WASC model is given by
Y
t
,
t
(, ) = E
_
e
i,Y
t+
[
t
, Y
t
(11)
= exp Tr [A()
t
] +i, Y
t
) +c() , (12)
where = (
1
, . . . ,
n
)
R
n
and the deterministic function A(t) M
n
is as
follows:
A() = A
22
()
1
A
21
() , (13)
with
_
A
11
() A
12
()
A
21
() A
22
()
_
= exp
_
M +Q
2Q
Q
1
2
_
(i)(i)
n
j=1
i
j
e
jj
_
_
M
+i
Q
_
_
.
(14)
6
The function c() can be obtained by direct integration, thus giving:
c() =
2
Tr
_
log (A
22
()) +M
+i(
Q)
+Tr
_
i
. (15)
The characteristic function of the Wishart process is dened as:
t
(, ) = E
_
e
iTr[
t+
]
[
t
, (16)
where M
n
.
Proposition 2.2. (Da Fonseca, 2006) Given a real symmetric matrix D, the
conditional characteristic function of the Wishart process
t
is given by:
t
(, ) = E
_
e
iTr[
t+
]
[
t
, Y
t
= exp Tr [B()
t
] +C() , (17)
where the deterministic complexvalued functions B() M
n
(C
n
), C() C are
given by
B() = (iB
12
() +B
22
())
1
(iB
11
() +B
21
()) (18)
C() =
2
Tr
_
log(iB
12
() +B
22
()) +M
,
with
_
B
11
() B
12
()
B
21
() B
22
()
_
= exp
_
M 2Q
Q
0 M
_
.
What is more, both these characteristic functions can be derived with respect
to and the elements of M, Q and , using the results of Daleckii (1974), on
the derivative of a matrix function. We provide detailed calculations of these
derivatives in the Appendix.
3 Possible estimation strategies
In this section, we review the possible estimation strategies for the WASC. We
discuss how similar processes have been estimated in special correlation cases,
using the Dynamic Conditional Correlation model introduced by Engle (2002)
to make the timevarying covariances observable. This methodology has been
applied by Buraschi et al. (2006) to a particular specication of the WASC model
corresponding to the choice = (1, 0)
t1
+Q
t1
R
t
= Q
t
1
Q
t
Q
t
1
,
with Q
t
= diag(
q
ii
) and
t
^(0, R
t
). Once again suitable constraints should
be imposed on and in order to guarantee a wellbehaved R
t
matrix.
As long as the DCC model provides consistent estimates of the timevarying
variancecovariance matrices, it is possible to estimate M, Q and consistently,
by using the unconstrained rstorder method of moments estimator presented
in Gourieroux et al. (2004). In our case, such an approach is precluded by the
presence of : in fact, the DCC does not allow for correlation between the returns
vector and the covariance matrix, while this correlation is nonzero in the WASC
model and is parametrized by . To see why, recall that in the DCC model the
variances are assumed to follow a GARCH(1,1) specication. Thus, the dynamic
of the i
th
component is:
r
i
t
=
_
h
i
t
i
t
, (21)
i
t
N(0, 1) (22)
h
i
t
=
0
+
1
h
i
t1
+
2
(r
i
t1
)
2
. (23)
9
When computing the skew, we obtain:
Cov
t1
(r
i
t
, h
i
t+1
h
i
t
) =
2
h
3
2
t
E
t1
[(
i
t
)
3
] = 0. (24)
Thus, the variances obtained with a DCC model have no correlation with the
returns. A similar computation can be perform to show that the conditional co
variance between the log of the asset price and the correlation process in the DCC
framework is equal to 0. These two key features preclude any DCCbased estima
tion of in the WASC. This problem is wellknown in nancial econometrics, at
least since Nelson and Foster (1994). We report in Figure 1 the empirical distri
bution of the skew and estimators with DCCbased timevarying covariances,
when the true (simulated) datagenerating process is the WASC, illustrating this
correlationestimation problem. Thus DCCbased estimates cannot be used. One
interesting exception is the model presented in Cappiello et al. (2006). However,
in this case, as in the DCC case, we fear modellinked measurement errors: in
particular, we cannot quantify how well these models yield a consistent estima
tion of the skew.
Finally, it is noteworthy to mention that the DCC cannot be used as a discretiza
tion of the continuous time WASC model for estimation purposes. As presented
in Nelson and Foster (1994) in the univariate case, the use of a GARCH process
to estimate a square root diusion process lead to an estimator that is expected
to have a very low eciency. The WASC process being a matrix square root
process, we would face similar if not worst diculties by undertaking such a
strategy in our multidimensional settings. Coming back to DCC, this model is
usually estimated in a twostage framework as volatilities and correlations can be
estimated separately. Clearly, the use of such an estimation strategy is dubious
for the WASC model since volatilities and correlations dwell on the same set of
parameters. We develop this point in the empirical results section as we detail
the dynamics of volatility and correlation in the twodimension case. On the basis
of these arguments, the use of the DCC to estimate the WASC should clearly be
discarded as a possible estimation strategy. For these reasons, we present in the
next subsection a converging discretization of the WASC process.
3.3 Approximation of the WASC model by Stochastic Dif
ference Equations
Another way to estimate continuous time processes consists in building a discrete
time model that converges towards the continuous time model, when the time lag
goes to zero. A methodology of this kind has already been applied to stochastic
volatility models by Nelson (1990). In this section, we build up a discretetime
model of this kind that converges in probability towards the WASC model.
Following the seminal paper of Nelson (1990), we build up a stochastic dierence
equations approximation of the WASC. Our approach relies on the fact that
a Wishart process with integer Gindikin value can be written as a square of a
10
OrnsteinUhlenbeck matrix process. On this point, see Bru (1991). This property
has been used by Gourieroux et al. (2004) to construct an estimator of a pure
Wishart process. Thus, there exists a matrixsized stochastic process X
t
in M
n
such that
t
= X
t
X
t
and whose dynamics are given by
dX
t
= Mdt +dN
t
Q, (25)
with N
t
; t 0 a n matrix Brownian motion and
Q the rotation of Q,
so that
Q is symmetric. What is more, the link between N and W is known
from Bru (1991). Thus, once the link between Z and N has been found (see
the Appendix), we are able to specify the WASC with the following stochastic
dierential equation:
_
dY
t
=
_
+
1
2
Vec((X
X)
ii
)
_
dt +
XdZ
t
dX
t
= Mdt +
_
diag(dZ
t
)
t
+ diag(d
Z
t
)
t
+ diag(d
Z
t
)
t
+ diag(d
Z
t
)
t
_
Q,
(26)
with Z,
Z,
Z and
Z independent Brownian motions. ,
,
and
are func
tions of the (rotated) parameter and of X
t
. Using this result and the ones
in Nelson (1990), we can provide a discrete time model converging towards the
WASC. From the Hermite polynomials theory we can build from
k
; k N
some sequences of uncorrelated normalized Gaussian variables, due to the fact
that E[H
i
(U)H
j
(U)] =
ij
c
i
with ^(0, 1), where
ij
is the Kronecker function
and c
i
= i!. By taking the limit, these sequences will converge to independent
Brownian motions, thus extending to a dimension greater than two the theorem
3.2 of Nelson (1990). We illustrate such converging sequences in the following
lemma, whose proof is provided in the Appendix.
Lemma 3.1. Let t
k
; k N be such that t
i
t
i1
= . Denote by (Y
t
k
, X
t
k+1
)
the solution of the stochastic dierence equations
_
Y
t
k
Y
t
k1
=
_
+
1
2
Vec((X
t
k
)
t
k
)
ii
_
+
_
X
t
k
_
t
k
t
k
X
t
k+1
= X
t
k
+M +
N
t
k
,
where
t
k
; k N are i.i.d ^(0, I
n
),
N
t
k
is given by
N
t
k
=
_
H
1
(
1
t
k
) 0
0 H
1
(
2
t
k
)
_
t
k
c
1
+
_
H
2
(
1
t
k
) 0
0 H
2
(
2
t
k
)
_
t
k
c
2
+
_
H
3
(
1
t
k
) 0
0 H
3
(
2
t
k
)
_
t
k
c
3
+
_
H
4
(
1
t
k
) 0
0 H
4
(
2
t
k
)
_
t
k
c
4
,
with c
i
= E[H
i
(U)
2
], U ^(0, 1) and H
i
(x) is the i
th
Hermite polynomial.
The process (Y
t
k
, X
t
k+1
) converges weakly to (Y
t
, X
t
) as t 0, with (Y
t
, X
t
)
dened previously.
Even though this approach is attractive, insofar as it makes the conditioning
easier, we will not pursue this direction. As already pointed out by Nelson (1990),
11
it remains to prove that the discrete approximation scheme will provide consistent
estimates of the continuous time model parameters. The question of consistency
and eciency, when dealing with unobservable components in the dynamics, is
of tremendous importance. We presented the previous discretization in order to
cast some light on the possibility of estimating a discrete version of the WASC,
without using the GARCHlike model, because of the need to estimate the
vectorsized parameter. Nevertheless, we have little control on the impact of the
discretization error brought about by such a procedure. Thus, instead of following
the approximation procedure, we prefer to turn our attention toward the use of
the exponential ane characteristic function of the WASC process, for which
we have wellestablished results related to Generalized Method of Momentbased
estimation procedures.
4 Spectral GMM in the WASC setting
Recent papers presented estimation methodologies using the empirical character
istic function as an estimation tool, since this function has a tractable expression
for many continuous time processes. In this section, we present how to estimate
the WASC in this framework, building on the approaches developed in Chacko
and Viceira (2003) and Carrasco et al. (2007).
The usual way to present the generalized method of moments based on spectral
moment conditions unfold as follows. Let h
t
be the conditional moment condition
such that
h
t
= e
iw,Y
t+
Y
t
X, (27)
with the notations developed earlier. X is a non stochastic function of the process
parameters. Let g(Y
t
) be the instruments. We need to identify X such that the
following relation holds:
E[h
t
g(Y
t
)] = 0 E[e
iw,Y
t+
Y
t
g(Y
t
)] E[Xg(Y
t
)] = 0. (28)
As what can be found in Chacko and Viceira (2003), the usual way to proceed
consists in setting
X = E[e
iw,Y
t+
Y
t
[Y
t
]. (29)
In the WASC case, X is thus set to be equal to :
X = e
c()
E[e
A(),
t
[Y
t
] = e
c()
0
(t, iA()), (30)
with c() dened in equation (15), A() dened in equation (13) and
t
(.) de
ned in equation (17). Clearly, in such a case equation (28) has no reason to
be equal to zero. The relation only holds if (Y
t
)
tN
and (
t
)
tN
are independent
processes or if g(Y
t
) = 1. The autonomous behavior of the matrix volatility is
not equivalent to having an independent process. Since in the WASC case this
12
independence is false, we imposed the instrument value to be 1, as in Chacko and
Viceira (2003) (see page 272). A spectral GMM approach cannot be undertaken
otherwise
1
. This setting stems from the fact that we integrate the volatility out
when computing X. Could we use
t
as an instrument, a more general form of
instruments would be readily used.
In order to increase the eciency of our estimates, we use a continuum of moment
conditions, as presented Carrasco et al. (2007). Note that the fact that we set the
instruments to be equal to 1 naturally prevents us from reaching the ML eciency
of CGMM estimates of Carrasco et al. (2007). Anyway, Y
t
conditionally upon its
past is no longer a Markov process, since the covariance matrix is unobservable.
ML eciency cannot be reach for nonMarkov process: the special instruments
chosen here does not necessarily jeopardize the estimation results.
Let now
h
t
(.) be the sample mean of the moment condition, that is a function
from R
2n
to C. In an innite conditions framework, Carrasco et al. (2007) showed
that the objective function to minimize is:
= arg min
K
1/2
h(), (31)
where K is the covariance operator, that is the counterpart of the covariance
matrix in nite dimension as in standard GMM approach and . is the weighted
norm
f
2
=
_
R
n
_
R
n
f()f()()d
where denotes any probability measure. As in Carrasco et al. (2007), we chose it
to be the normal distribution. Carrasco et al. (2007) showed that the covariance
operator K can be written as follows:
Kf() =
_ _
k(, )f()()d
where the function k is the so called kernel of the integral operator K and is
dened by:
k(, ) =
+
j=
E
0
_
h
t
(;
0
)h
tj
(;
0
)
_
.
Since our approach is nested within the Carrasco et al. (2007)s, we now thor
oughly follow their settings. Our approach can also be related to the methodology
presented in Rockinger and Semenova (2005). In order to construct an estimator
of the covariance operator, Carrasco et al. (2007) proposed a twostep procedure.
The rst step consists in nding:
1
= arg min
h(). (32)
1
We thank Marine Carrasco for pointing out this fact.
13
Since h
t
is not a martingale dierence series, the second step consists in estimating
the kernel k as follows:
k(
s
,
r
,
v
,
w
) =
T
T q
T1
j=T+1
_
j
S
T
_
T
(j), (33)
with
T
(j) =
_
1
T
T
t=j+1
h
t
(
s
,
r
,
1
)h
tj
(
v
,
w
,
1
), j 0
1
T
T
t=j+1
h
t+j
(
s
,
r
,
1
)h
t
(
v
,
w
,
1
), j < 0,
, (34)
where w(.) is any kernel satisfying some regularity conditions (see Carrasco et al.
(2007) Appendix A.6) and S
T
is a bandwidth parameter.
Once the covariance operator is estimated, the minimization in equation (31)
requires the computation of the inverse of K. Unfortunately, K has typically
a countable innity of eigenvalues decreasing to zero, so that its inverse is not
bounded. We need then to regularize the inverse of K, which can be done by
replacing K by a nearby operator that has a bounded inverse, due to the presence
of a penalizing term. Carrasco et al. (2007) used the Tikhonov approximation of
the generalized inverse of K. Let be a strictly positive parameter, then K
1
is
replaced by:
(K
)
1
= (K
2
+I)
1
K. (35)
As outlined in Carrasco et al. (2007), the choice of is important but does not
jeopardize the consistency of the estimates. Carrasco and Florens (2000) investi
gated an empirical method to select its value, and the optimal value for it should
represent a tradeo between the instability of the generalized inverse (for small
values of ) and the distance from the true inverse as increases. Furthermore
we found much more convenient to compute (K
)
1
using the Choleskys de
composition than the spectral decomposition: it is sucient for the evaluation of
equation (31) and avoids the numerically dicult problem of eigenvectors com
putation.
Under mild regularity conditions (conditions A.1. to A.5. in Carrasco et al.,
2007), it can be proved that the optimal CGMM estimator of is obtained by:
= arg min
(K
T
)
1/2
h
T
() (36)
and is asymptotically Normal with
T(
T
0
)
L
^
_
0, (E
0
(
h), E
0
(
h))
K
)
1
_
, (37)
as T and T
a
5/4
T
go to innity and goes to zero. (
11
t
. In the WASC
framework, individual parameters can hardly be interpreted on their own: on the
contrary, combinations of these parameters have standard nancial interpreta
tions, such as the meanreverting parameter or the volatility of volatility. Now,
we review the computation of these quantities.
For the rst asset, the quadratic variation of the volatility can be computed as
follow:
d
11
,
11
)
t
= 4
11
t
(Q
2
11
+Q
2
21
)dt, (38)
for the rst asset. Therefore the rst column of Q parametrizes the volatility of
volatility of the rst asset. Similar results can be obtained for the second asset.
Then, as presented in Section 2,
corr
_
dY
1
, d
11
_
=
Q
11
1
+Q
21
2
_
Q
2
11
+Q
2
21
, (39)
where corr(.) is the correlation coecient. As already mentioned, the short term
behavior of the smile and the skewness eect heavily depend on the correlation
structure given by the vector . If Q and are such that this quantity is negative,
then the volatility of S
1
will rise in response to negative shocks in returns of this
asset. We expect this correlation to be large and negative, in order to account
for the large skewness found in nancial datasets.
The Gindikin coecient insures the positiveness of the Wishart process. What
is more, an increase of it will shift the distribution of the smallest eigenvalue
to positive values. Thus, this parameter can be interpreted as a global variance
shift factor. From equations (3) and (4), if is multiplied by a factor , the
long term covariance matrix
22
t
_
11
t
12
t
_
+. . . , (40)
where
12
t
is the instantaneous correlation between the logreturns of the two
assets. Thus, the mean reverting parameter for
11
is a combination of the ele
ments of M. What is more, this drift term is made of two parts: a deterministic
part (2M
11
) and a stochastic correction (2M
12
22
t
11
t
12
t
), linked to the joint dy
namics of both assets. Thus, the drift term of
11
t
is inuenced by one of the
odiagonal elements of M. This feature cannot be replicated by most of the mul
tivariate GARCHlike models. We can perform similar calculations for
12
t
and
22
t
. These quantities can then be used to compare the half life of the variances
and covariance processes and thus evaluate their relative persistence in nancial
markets.
The instantaneous correlation between assets has also a closed form expression:
d
12
t
=
_
A
t
_
12
t
_
2
+B
t
12
t
+C
t
_
dt +
_
1 (
12
t
)
2
(...)d(Noise
t
) (41)
with A
t
, B
t
, C
t
recursive functions of
11
t
,
22
t
and the model parameters. We
present the drift coecients and the diusion term in the Appendix. The drift
associated to the correlation is quadratic, and the linear term has a negative co
ecient B
t
< 0, thus presenting the typical mean reverting behavior of
12
t
(at
least around zero where the quadratic part is negligible). The linear part can
thus be used to analyze the persistence of the correlation and its meanreverting
characteristics, during low correlation periods. When the absolute value of the
correlation is higher, the quadratic part of the drift get the upper hand and the
correlation process looses most of its persistence. By comparing the values of
B
t
and A
t
, when can thus compare the correlation behavior during low and high
correlation cycles. This information has not been documented until now, whereas
it is important to understand the joint behavior of nancial assets.
The WASC model can also be used to investigate potential contagion eects in
nancial markets. By computing the correlation between the correlation process
and the returns, we can discuss under which condition the model is able to display
16
an asymmetric correlation eect
2
. Asymmetric correlation eect leads correlation
to go up whilst returns are getting down. As already noticed in Da Fonseca et al.
(2006), we have:
dY
1
,
12
)
t
=
11
t
22
t
(1
_
12
t
_
2
) (Q
12
1
+Q
22
2
)
. .
Sign of asset 2 skew
. (42)
Thus, the sign of the skews determines the one of the covariance between correla
tion and returns. Were the skew to be negative and the model would also display
increases in the correlation following negative returns. Thus, the WASC model is
also able to display an asymmetric correlation eect, whose sign is driven by the
skewness associated to the returns. Since the asset returns are negatively cor
related to their own volatility (leverage eect), we thus expect volatilities to be
positively correlated to correlation: negative returns periods correspond to both
higher correlation and higher volatility periods. In fact, simple computations
given in Appendix lead to
d
12
,
11
_
t
=
11
t
22
t
_
1
_
12
t
_
2
_
Q
12
_
Q
11
+
Q
22
_
dt.
where
Q is the symmetric positive denite matrix associated to the polar decom
position of Q
3
. A positive value for
Q
12
would mean that the WASC model is able
to accomodate stylized eects of the type mentioned earlier. Due to the increase
in the drift term of the correlation dynamics, situation of this kind are expected
not to last for long.
We now turn our attention toward a series of Monte Carlo experiments, so as to
investigate the empirical performance of the chosen estimation strategy.
5.2 Monte Carlo study
Following Carrasco et al. (2007), we present the results of a Monte Carlo study
of the CGMM estimation methodology applied to the WASC. We rst present
the technical details of the simulation and then we review the results obtained.
For the ease of the presentation, we restrict to the twoassets case. The parame
2
On asymmetric correlation eects, see Roll (1988) and Ang and Chen (2002).
3
Any invertible matrix Q can be uniquely written as the product of a rotation matrix and
a symmetric positive denite matrix
Q, see the Appendix.
17
ters used in the simulation are the following:
0
=
_
0.0225 0.0054
0.0054 0.0144
_
(43)
M =
_
5 3
3 5
_
(44)
=
_
0.3 0.4
(45)
Q =
_
0.1133137 0.03335871
0.0000000 0.07954368
_
(46)
= 15. (47)
The Q matrix is obtained by inverting the relation that links Q to M,
and :
Q
Q =
1
_
M
_
. (48)
This ensure the stationarity of the correlation process. When Q is selected arbi
trary and given the mean reverting property of
t
, the rst part of the simulated
sample will be tainted by the collapse of the process toward its long term average.
Situations of this kind should be discarded. The Figure 2 presents a simulated
path for both volatilities and correlation, using the previous parameter values.
The gure displays meanreverting dynamics for each of these moments.
The Figure 3 shows the characteristic function used in the spectral GMM method
used in the paper, as presented in equation (29). The grid used for the numerical
integration of the objective function ranges on the real line from 300 to 300.
We used Gaussian kernels with appropriate variance parameter to maintain as
much information as possible. The integral is computed numerically using the
Trapezoidal Rule that seemed to performed well over the simulated dataset. The
objective function is minimized using a simulated annealing method, as described
in Belisle (1992).
We present the results of dierent Monte Carlo experiments. Each of them comes
out after 1000 iterations, but they dier by the length of the simulated sample
and the sampling frequency: daily, weekly and monthly. For each sampling fre
quency, we used two dierent samples, one of which contains 500 observations
and the other one 1500 observations. The Table 7.3 presents the Mean Bias and
the Root Mean Square Error (RMSE) obtained. We did not reported the median
bias insofar it was close to the median bias, thus indicating that the estimators
have a symmetric empirical distribution.
The results can be analyzed as follows. The Monte Carlo results obtained for Q,
and show that an increase in the sample depth globally results in a reduction
of the variance of the estimates. The bias obtained are small and not signica
tive. For the weekly frequency,
1
displays a noticeable dierence as the variance
of the estimate grows with the sample size. This feature will have to be consid
ered when analyzing the real datasetbased estimation results. The M parameter
18
also presents this variance increase feature. However, this behavior is not very
surprising: a large number of articles emphasize the diculties involved by the
estimation of the mean reverting parameter in continuous time diusions (see e.g.
Gourieroux and Monfort, 2007). The Monte Carlo results indicate that this mean
reverting parameter is estimated with less volatility with daily series. What is
more, diagonal elements (resp. odiagonal elements) of M are estimated with a
small positive (resp. negative) bias and thus may be underestimated (resp. over
estimated) when working with a realtime dataset. Finally, the correlation vector
displays a remarkably small bias and small RMSE for the daily datasets, even
in the small sample version. This fact is somewhat constant for each sampling
frequency. This point is important for the WASC model, given that we are in
terested in the analysis of the ne correlation structure implicit in asset dynamics.
We now detail the empirical results obtained with stock indexes.
5.3 Estimation on stock indexes
In this last subsection, we present the empirical results obtained when estimating
the WASC using the CGMM method on a real dataset. We used the following
stock indexes: SP500, FTSE, DAX and CAC 40. For each stock, the time series
starts on January 2nd 1990 and ends on June 30th 2007. This period excludes
the 1987 crash and the subprime crisis. It nonetheless includes a lot of nancial
turmoils, as pointed in Rockinger and Semenova (2005). The table 2 presents
the descriptive statistics for the sample used in the estimation, at a weekly sam
pling frequency. We used daily and weekly time series. We discarded the use of
monthly ones since the sample would be far too small. In many articles devoted to
the estimation of continuous time models, the change in the sampling frequency
usually leads to an interesting analysis of the subtle dynamics of nancial mar
kets (see e.g. Chacko and Viceira, 2003). Since the characteristicfunction based
estimators do not suer from discretization errors, we can actually use any sam
pling frequency. Like in Carrasco et al. (2007), = 0.02 were found to perform
well. The integration grid is the same as the one used for the previous simulation
exercises. We chose to use the Bartlett kernel for the GMM covariance matrix
estimation, following the procedure presented in Newey and West (1994).
For numerical sake, we focus again on a twoassets case (n = 2). We estimated
the parameters driving the WASC process for the following couples of indexes:
(SP500,FTSE), (SP500,DAX), (SP500,CAC), (DAX/CAC), (DAX,FTSE) and
(FTSE/DAX). This way, we will be able to compare the characteristics specic
to individual stock while estimated with each of the others. For example, we will
be able to compare the volatility of volatility of the SP500, when it is estimated
with the DAX, CAC and FTSE as a second asset. It will highlight the impact of
joint dynamics on idiosyncratic behaviors, which has hardly be documented until
now.
The estimation results are presented in table 3 for the daily results and in table
19
4 for the weekly results. Most of the estimates are signicative up to a 5 or 10%
risk level. What is more, in the weekly observation case, the size of the sample is
strongly reduced and so is the eciency of the estimation method. Nevertheless,
the estimation results yield interesting information both about the WASC process
and the dynamics of the stock indexes.
As presented in the previous subsection, it is dicult to compare the individual
parameters and we will thus focus on combinations of these parameters most of
which are comparable with the ones of the Heston (1993) model.
For the estimated parameters presented in Table 3, the associated volatility of
volatility are presented in Table 5. The estimation of this quantity is essential
to test the ability of the model to capture nancial market features: as pointed
out in Chacko and Viceira (2003), this parameter controls the kurtosis of the
underlying process. Several remarks can be made. First, the global results match
what is generally expected from stock indexes. Such markets are known to lead to
a volatility of volatility ranging from 5% to 25%. Second, the results obtained for
the SP500 are remarkably stable when the second asset changes at least for the
daily sample: it ranges from 14.6% to 24.4%, thus matching the results obtained
in Eraker et al. (2003). Still, it is below the estimates obtained in Chacko and
Viceira (2003) and Rockinger and Semenova (2005): this may be explained by
the fact that the model that is estimated here is multivariate, whereas existing
attempts to capture stochastic volatility has been made in a univariate framework.
Da Fonseca et al. (2006) showed that the correlation between the volatility of each
stock is non 0 insofar as
d
11
,
22
) =
12
dt. (49)
Hence, whenever
12
is positive, the WASC model is able to model volatility
transmission phenomena among assets. It is noteworthy to remark that these
results are globally stable across the datasets and close to the existing results.
Third, when the sampling frequency reduces, the volatility of volatility parame
ter globally increases. The few lacks of consistency for this fact may be due to
the fact that the number of observations in the weekly dataset is far below the
one used in the daily dataset. These results are dierent from those obtained in
Chacko and Viceira (2003). However, this is in line with what is observed for the
volatility of the logreturns when reducing the sampling frequency. This diver
gence may also be explained by the eect of the correlation between variances
that cannot be mimicked in a Hestonlike framework.
Now, let us discuss an important parameter for the specication of the WASC
model, that is the correlation between the returns and the volatility. We already
mentioned that this parameter is essential to have a model that is consistent with
many stylized facts, such as negative skewness and thus skewed implied volatility
surfaces. It can be computed using its expression given in equation (39). The
Table 6 displays the results obtained.
20
This time, we have results that are comparable to the one obtained in the existing
literature, and especially for the SP500. The correlation for this index is reported
in the rst line of the previous Table. In the literature, it actually ranges from
0.27 (Rockinger and Semenova, 2005) to 0.62 (Chacko and Viceira, 2003), which
is close to what is obtained here. The parameter values obtained for the CAC,
DAX and FTSE indexes are not surprising either, since their sign is negative.
The main problem here lies in the instability of the parameter for the dierent
estimation involved, when comparing both the sample frequency and the couple
of indexes that are estimated. The change in sampling frequency does not lead
to a similar behavior across the datasets: depending upon the couples and the
sampling frequency, the skewness in the dataset can considerably dier, underlin
ing the fact that correlation processes implicit in nancial markets are complex.
Beyond the remarks made in the previous paragraph on the importance of the
dataset depth, we also emphasize that the computation of this parameter dwells
on the inverse of the square root of a quantity that is small. In this situation,
the inverse of something small can be found to be very variable: any error in the
estimation of Q
11
and Q
21
will have a strong impact on
1
Q
2
11
+Q
2
21
. Thus, this
skewness quantity must be cautiously interpreted. Last but not least, since the
skews are negative, the tted WASC models also display asymmetric correlation
eects: negative returns are likely to be followed by a higher correlation between
the two assets.
We now turn our attention toward the mean reverting matrix M. For SP500,
CAC, DAX and FTSE, we nd the same structure for the matrix M. They are
denite negative thus ensuring a mean reverting behavior for the Wishart process
and have positives o diagonal terms. As presented in the previous subsection,
the drift can be decomposed into two dierent part: an idiosyncratic part (de
noted
1
in the tables) and a joint part (denoted
2
in the tables). For univariate
stochastic volatility models, the estimation results usually lead to an estimate
of =
1
+
2
, that is the sum of the two preceding elements. Thanks to the
complexity of the WASC process, we are now able to disentangle and analyse
these two dierent elements. The estimation results are reported in table 8. The
values should be compared with the mean reverting value of the Heston. We
are close to Rockinger and Semenova (2005) results who found 6.3352 (see their
Table 1) for the S&P500, even though their results are obtained on a dierent
sample and using daily data. However, when analysing
1
and
2
, we nd that the
idiosyncratic mean reverting component is always higher than the usual Heston
parameter. This idiosyncratic element is dampened by the negative joint mean
reverting component: its negativity is to be related to the negative nondiagonal
elements of the estimated M matrices. Again, when the sampling frequency
changes, each of these values vary, suggesting that the mean reverting parame
ter associated to the volatility strongly depends on the sampling frequency, as
pointed out in Chacko and Viceira (2003). Globally, the associated half lives are
around one month, which is a realistic value.
21
As presented in the previous section, it is possible to perform similar computa
tions for the drift term of the stochastic correlation. This drift is a non linear
function of
12
t
and the usual comments have to be adapted. We present in gure
4 and 5 the instantaneous variation of
12
t
as a function of
12
t
, highlighting the
contribution of the quadratic term when the correlation gets very high that
is during crisis period. Our results suggest that the correlation process is much
more persistent than volatility when the correlation is below its long term level,
since in such a case its mean reverting parameter can be reduced to (B
t
): in
this situation, the half life is around 2 months, which is again realistic. When
correlation is high, the quadratic term gets higher and the persistence goes down,
since A
t
is added to B
t
as both these elements are negative. This is consistent
with what is empirically observed during nancial market crises: the correlation
gets very high on a very short period, to nally go back to its long term behavior
rapidly. The aforementioned gures display reaction functions of this kind, un
derlining the ability of the WASC model to encompass this standard feature of
nancial markets.
Another quantity of importance is [[[[, since the WASC can be seen as a general
ization of the processes proposed in Gourieroux and Sufana (2004) and Buraschi
et al. (2006), with a more complex correlation structure. Since the WASC model
is only dened up to a rotation matrix, the model presented in Buraschi et al.
(2006) encompasses any correlation structure that satises [[[[ = 1. Testing
such an assumption is thus of a tremendous importance to judge whether the
complexity of the WASC is empirically justied. The table 10 presents the norm
of this vector parameter. Each of the estimated value strongly diers from 1,
suggesting the general correlation structure imposed in Da Fonseca et al. (2006)
is empirically grounded.
As presented earlier, a contagion eect can be handled by the WASC through a
positive value for
Q
12
. In table 11 we report the estimated values for this param
eter. We found positive values for all couples of indexes as expected. Therefore,
the estimated WASC model is able to detect the existence of potential contagion
eects in the dataset. As mentioned earlier, these ndings may be due to the fact
that the dataset includes several nancial crises, periods during which dramatic
contagion eects are expected.
6 Conclusion
In this paper we investigated the estimation of a new continuous time model:
the Wishart Ane Stochastic Correlation model, presented in Da Fonseca et al.
(2006). After having presented the problem that arise when trying to estimate
a discrete version of this model, this paper proposes to estimate the process us
ing its exponential ane characteristic function. The estimation method uses a
continuum of spectral moment conditions in a GMM framework. After a pre
22
liminary Monte Carlo investigation of the estimation methodology, we show that
realdataset estimation results bring support to the WASC process. First, the
empirical results are comparable to those obtained in the literature (when com
parable). Second, the general correlation structure of the WASC casts light on
notsowell documented features of international equities, allowing us to discuss
for example the persistence of the correlation process, contagion eects or asym
metric correlation eects. Third, the generality of the correlation structure is
not rejected by the dataset, bringing empirical support to the model presented
in Da Fonseca et al. (2006).
References
Ang, A. and Chen, J. (2002). Asymmetric Correlations of Equity Portfolios.
Journal of Financial Economic, (63):443494.
Asai, M., McAleer, M., and Yu, J. (2006). Multivariate Stochastic Volatility: a
Review. Econometric Review, 25(23):145175.
BarndorNielsen, O. and Shephard, N. (2002). Econometric Analysis of Realised
Volatility and its Use in Estimating Stochastic Volatility Models. Journal of
the Royal Statistical Society, 63:253280.
BarndorNielsen, O. E. and Shephard, N. (2004). Econometric analysis of real
ized covariation: High frequency based covariance, regression, and correlation
in nancial economics. Econometrica, 72(3):885925.
Bathia, R. (2005). Matrix Analysis. Graduate Texts in Mathematics, Springer
Verlag.
Belisle, C. J. P. (1992). Convergence Theorems for a Class of Simulated Annealing
Algorithms. Rd J Applied Probability, 29:885895.
Bru, M. F. (1991). Wishart Processes. Journal of Theoretical Probability, 4:725
743.
Buraschi, A., Porchia, P., and Trojani, F. (2006). Correlation risk and optimal
portfolio choice. Working paper, SSRN908664.
Cappiello, L., Engle, R. F., and Sheppard, K. (2006). Asymmetric Dynamics
in the Correlations of Global Equity and Bond Returns. Journal of Financial
Econometrics, 4(4):537572.
Carrasco, M., Chernov, M., Florens, J.P., and Ghysels, E. (2007). Ecient
Estimation of Jump Diusions and General Dynamic Models with a Continuum
of Moment Conditions. Journal of Econometrics, (140):529573.
Carrasco, M. and Florens, J. (2000). Generalization of GMM to a Continuum of
Moment Conditions. Econometric Theory, (16):797834.
23
Chacko, G. and Viceira, L. M. (2003). Spectral GMM Estimation of Continuous
Time Processes. Journal of Econometrics, 116(12):259292.
Da Fonseca, J., Grasselli, M., and Tebaldi, C. (2005). Wishart MultiDimensional
Stochastic Volatility, to appear as: A Multifactor Volatility Heston Model in
Quantitative Finance. Mimeo Ecole Superieure dIngenieurs Leonard de Vinci,
RR31.
Da Fonseca, J., Grasselli, M., and Tebaldi, C. (2006). Option Pricing when
Correlations are Stochastic: an Analytical Framework. To appear in Review of
Derivatives Research.
Daleckii, J. (1974). Dierentiation of NonHermitian Matrix Functions Depending
on a Parameter. AMS Translations, 47(2):7387.
Daleckii, J. and Krein, S. (1974). Integration and Dierentiation of Functions of
Hermitian Operators and Applications to the Theory of Perturbations. AMS
Translations, 47(2):130.
Donoghue, W. J. (1974). Monotone matrix functions and analytic continuation.
SpringerVerlag.
Due, D. and Singleton, K. (1993). Simulated Moments Estimation of Markov
Models of Asset Prices. Econometrica, 61:929952.
Engle, R. (2002). Dynamic Conditional Correlation: A Simple Class of Multivari
ate Generalized Autoregressive Conditional Heteroskedasticity Models. Journal
of Business & Economic Statistics, 20(3):33950.
Eraker, B., Johannes, M., and Polson, N. (2003). The Impact of Jumps in Volatil
ity and Returns. The Journal of Finance, 58(3):12691300.
Faraut, J. (2006). Analyse sur les groupes de Lie. Calvage & Mounet.
Gourieroux, C. (2006). Continuous Time Wishart Process for Stochastic Risk.
Econometric Review, 25(23):177217.
Gourieroux, C. and Jasiak, J. (2001). Financial Econometrics. Princeton Uni
versity Press.
Gourieroux, C., Jasiak, J., and Sufana, R. (2004). The Wishart Autoregressive
of Multivariate Stochastic Volatility. Working paper, University of Toronto,
(200432).
Gourieroux, C. and Sufana, R. (2004). Derivative Pricing with Multivariate
Stochastic Volatility: Application to Credit Risk. Les Cahiers du CREF 0409.
Hall, B. C. (2003). Lie Groups, Lie Algebras, and Representations: An Elemen
tary introduction. Graduate Texts in Mathematics 222, SpringerVerlag.
24
Harvey, A., Ruiz, E., and Shephard, N. (1994). Multivariate stochastic volatility
models. Review of Economic Studies, 61:247264.
Heston, S. (1993). A ClosedForm Solution for Options with Stochastic Volatility
with Applications to Bond and Currency Options. The Review of Financial
Studies, 6(2).
Jiang, G. J. and Knight, J. L. (2002). Estimation of ContinuousTime Processes
via the Empirical Characteristic Function. Journal of Business & Economic
Statistics, 20(2):198212.
Lo, A. W. (1988). Maximum Likelihood Estimation of Generalized Ito Processes
with Discretely Sampled Data. Econometric Theory, 4:231247.
Nelson, D. B. (1990). ARCH models as diusion approximations. Journal of
Econometrics, 45:738.
Nelson, D. B. and Foster, D. P. (1994). Asymptotic Filtering Theory for Univari
ate Arch Models. Econometrica, (1):141.
Newey, W. K. and West, K. D. (1994). Automatic lag selection in covariance
matrix estimation. Review of Economic Studies, 61(4):63153.
Rockinger, M. and Semenova, M. (2005). Estimation of JumpDiusion Process
via Empirical Characteristic Function. FAME Research Paper Series rp150,
International Center for Financial Asset Management and Engineering.
Roll, R. (1988). The International Crash of October, 1987. Financial Analysts
Journal, (SeptemberOctober):1935.
Singleton, K. (2001). Estimation of Ane Pricing Models Using the Empirical
Characteristic Function. Journal of Econometrics, (102):111141.
Singleton, K. J. (2006). Empirical Dynamic Asset Pricing: Model Specication
and Econometric Assessment. Princeton University Press.
25
7 Appendix
7.1 Computing the gradient
The gradient of the characteristic function is needed to study the asymptotic
distribution of the estimates but also in the optimization process underlying the
estimation procedure. Therefore we turn our attention to the dierentiation of
matrix function depending on a parameter. We illustrate the theoretical frame
work with the characteristic function of the assets log returns and we give without
technical details the results for the forward characteristic function needed in our
empirical study. We mainly rely on the work of Daleckii (1974) for the general
case (i.e. in the nonHermitian matrix case) and to Daleckii and Krein (1974),
Donoghue (1974) and Bathia (2005) for the Hermitian matrix case.
Let us rst state some basic results on linear algebra. Denote by
i
; i = 1..n
the set of eigenvalues of a matrix X M
n
and m
i
the multiplicity of
i
as a
root of the characteristic polynomial of X. Dene L
i
= Ker(X
i
I) and P
i
the
projection operator from C
n
onto L
i
, then we have
n
i=1
P
i
= I. Dene also J
i
such that (X
i
I)P
i
= J
i
. The Jordan normal form of X if given by the well
known decomposition X =
n
i=1
(
i
P
i
+J
i
).
Let f be a function from M
n
into M
n
: the derivative of f at X in direction H,
denoted D
f,X
(H), is by denition f(X + tH) f(X) D
f,X
(H) = to(H)
and can be computed using the following formula Daleckii (1974):
D
f,X
(H) =
n
j
1
,j
2
m
j
1
1
r
1
=0
m
j
2
1
r
2
=0
1
r
1
r
2
r
1
+r
2
r
1
j
1
r
2
j
2
_
f(
j
1
) f(
j
2
)
j
1
j
2
_
P
j
1
J
r
1
j
1
HP
j
2
J
r
2
j
2
. (50)
Remark 7.1. Whenever
j
1
=
j
2
the term within the bracket should be replaced
by f
(
j
1
).
Remark 7.2. When X can be diagonalized then m
j
= 1 for each j and we are
lead to the very simple form
D
f,X
(H) =
n
j
1
,j
2
_
f(
j
1
) f(
j
2
)
j
1
j
2
_
P
j
1
HP
j
2
. (51)
If X is Hermitian then P
1
= P
if ,=
f
() if =
(52)
4
If p
i
is the i
th
eigenvector of X and q
i
is the i
th
row of P
1
then the projection operator
on L
i
is given by P
i
= p
i
q
i
5
Given two matrix of same size X
kl
and Y
kl
then X Y = X
kl
Y
kl
26
This formulation is well known and can be found for example in Donoghue (1974)
p. 79 and Bathia (2005) p123124.
The gradient of the characteristic function involves
Y
t
,
t
(, z) = (Tr(
A
t
) +
c())
Y
t
,
t
(, z).
From (13) and
(L
1
) = L
1
(
L)L
1
implied by the derivation of L
1
L = I
we conclude that
A() = A
22
()
1
(A
22
())A() +A
22
()
1
A
21
() ,
Therefore we are lead to the computation of
_
A
11
()
A
12
()
A
21
()
A
22
()
_
,
that is the derivative with respect to a parameter of a function of a matrix (in
this case the exponential function). In order to use formula (50) we specify in the
following table for each parameter of the WASC model the choice of the matrice
X and H. As usual e
l
; l = 1 . . . n resp. e
kl
; k, l = 1 . . . n stands for the
canonical basis of R
n
resp. M
n
(R) (the function f being f(x) = e
x
).
Parameter X H
M
kl
G
_
e
kl
0
0 e
kl
_
Q
ij
G
_
(e
kl
)
2(Q
e
kl
+ (e
kl
)
Q)
0 i
e
kl
_
l
G
_
Q
e
l
i
0
0 ie
l
Q
_
To fulll the analytical computation of the gradient we need the derivative of c()
with respect to a model parameter and particularly the term log A
22
().
c() =
2
_
Tr (
(log A
22
())) +Tr
_
+i
Q)
_
(53)
In order to apply (50) we just need to dene P
22
from M
2n
into M
n
such that
P
22
L = L
22
with
L =
_
L
11
L
12
L
21
L
22
_
. (54)
Then it is easy to see that
log A
22
() = D
log,A
22
()
(P
22
D
exp,G
(H)) (55)
Once again using (50) with the log function gives the result.
27
Remark 7.3. If f is the exponential function then we can also compute the
derivative of the exponential of a matrix using the BakerHausdors formula
(see e.g. Hall (2003) p. 71 formula (3.10) for the details)
e
G
= Dexp
G
(G), (56)
where ad
X
Y = [X, Y ] = XY Y X is the Lie bracket and Dexp
X
= e
X Ie
ad
X
ad
X
.
The empirical study is based on the forward characteristic function of assets log
returns dened by
0
(t, iA())e
c()
= exp Tr [B(t)
0
] +C(t) +c() (57)
with
B(t) = (A()B
12
(t) +B
22
(t))
1
(A()B
11
(t) +B
21
(t))
C(t) =
2
Tr
_
log(A()B
12
(t) +B
22
(t)) +tM
c() =
2
Tr
_
log(A
22
()) +M
+i(
Q)
0
(t, iA())e
c()
=
0
(t, iA())e
c()
(Tr(
B(t)
0
) +
C(t) +
c()))
where the matrix derivatives are given by
B(t) = (A()B
12
(t) +B
22
(t))(
A()B
12
(t) +A()
B
12
(t) +
B
22
(t))
1
B(t)
+ (A()B
12
(t) +B
22
(t))
1
(
A()B
11
(t) +B
21
(t)),
C(t) =
2
Tr(D
log,A()B
12
(t)+B
22
(t)
(
A()B
12
(t) +A()
B
12
(t) +
B
22
(t))).
This completes the analytical computation of the gradient.
7.2 WASC approximation
To prove the convergence of the discrete time process dened in Lemma 3.1 to
the WASC model we rst make some simplications.
Using the fact that Q GL(n, R)
6
we know that Q = K
Q with K O(n)
and
Q T
n
where O(n) stands for the orthogonal group
7
and T
n
is the set of
symmetric positive denite matrices. We refer to Faraut (2006) for basic results
on matrix analysis. The matrix Brownian motion being invariant under rotation
the dynamic can be rewritten as :
d
t
=
_
+M
t
+
t
M
_
dt +
_
t
d
W
t
Q+
Qd
W
t
_
t
. (58)
6
GL(n, R) is the linear group, the set of invertible matrices.
7
O(n) = g GL(n, R)[g
g = I
n
28
The associated correlation structure is = K
. Note that
Q
= Q
thus the
WASC models build upon (, M, Q, ) and (, M,
Q, ) have same characteristic
function. From now on we will work with the (, M,
Q, ) specication without
loss of generality.
To build a discrete time approximating scheme to the WASC it is tempting to
use an Euler scheme. Unfortunately a major diculty arises with the positive
ness of the Wishart process. One can not assure that the discrete approximating
scheme remains in T
n
. One way to solve the problem it to rewrite the Wishart
process as a product of OrnsteinUhlenbeck matrix processes thus ensuring posi
tiveness and to approximate, through an Euler scheme, the OrnsteinUhlenbeck
process. In the WASC the correlation between assets returns and their volatili
ties is controlled by the vector therefore we need to reformulate the correlation
between the assets and the OrnsteinUlhenbeck process. We turn to a more for
mal presentation. It is an extension to our framework of the strategy developed
in Gourieroux et al. (2004).
We make the following hypothesis on the Gindikin coecient: N and >
n 1. Under this hypothesis and following equation (5.7) of Bru (1991) there
exists a matrix stochastic process X
t
in M
n
such that
t
= X
t
X
t
and whose
dynamics are given by
dX
t
= Mdt +dN
t
Q, (59)
with N
t
; t 0 a n matrix Brownian motion. The matrix Brownian motion
in equation (58) is related to equation (59) through d
W
t
=
_
1
t
X
t
dN
t
. As
the correlation in the WASC is dened between Z
t
; t 0 and
W
t
; t 0, we
turn our attention toward the problem of specifying it with respect to N
t
; t
0. Starting from the fact that dZ
k
t
d
W
ij
t
=
j
ki
dt for i, j, k = 1, . . . , n
8
and
dening
kj
t
; k, j = 1, . . . , n such that dZ
k
t
dN
ij
t
=
kj
t
ki
we conclude that
kj
t
=
j
n
l=0
kl
t
X
kl
t
where
(
ij
t
)
i,j=1..n
=
_
(X
t
X
t
)
(
ij
t
)
i,j=1..n
=
_
ij
t
_
1
i,j=1..n
.
From now on we formulate the problem with n = = 2 without loss of generality.
Using the denition of
kj
t
we have :
_
_
dN
11
t
=
11
t
dZ
1
t
+
12
t
d
Z
1
t
+
_
1 (
11
t
)
2
(
12
t
)
2
d
Z
1
t
dN
12
t
=
12
t
dZ
1
t
11
t
d
Z
1
t
+
_
1 (
12
t
)
2
(
11
t
)
2
d
Z
1
t
dN
21
t
=
21
t
dZ
2
t
+
22
t
d
Z
2
t
+
_
1 (
21
t
)
2
(
22
t
)
2
d
Z
2
t
dN
22
t
=
22
t
dZ
2
t
21
t
d
Z
2
t
+
_
1 (
22
t
)
2
(
21
t
)
2
d
Z
2
t
,
with
Z
t
; t 0,
Z
t
; t 0 and
Z
t
; t 0 three independent Brownian motions
(two dimensional vector). Combining the dynamics of the assets (1), (59),
t
=
8
ij
is the Kronecker function.
29
X
t
X
t
and the decomposition of N
t
with respect to Z
t
,
Z
t
,
Z
t
and
Z
t
, we obtain
that (Y
t
= log(S
t
), X
t
) satisfy the stochastic dierential equations system
_
dY
t
=
_
1
2
Vec((X
X)
ii
)
_
dt +
_
X
t
X
t
dZ
t
dX
t
= Mdt +
_
diag(dZ
t
)
t
+ diag(d
Z
t
)
t
+ diag(d
Z
t
)
t
+ diag(d
Z
t
)
t
_
Q.
Finally, in order to build an approximation of the above diusions in the spirit
of Nelson (1990), it is enough to discretize the above SDE, thus leading to the
announced stochastic dierence equations. Since the process X
t
is driven by
more noises than Y
t
, we then use the Hermite polynomials in order to generate
independent noises. In the case of = 0 we recover the result of Gourieroux et al.
(2004).
7.3 Dynamics of the correlation process
In this Appendix we compute in the 2dimensional case the drift and the diusion
coecients of the correlation process
12
t
dened by
12
t
=
12
t
_
11
t
22
t
. (60)
The Wishart process
t
satises the following SDE:
d
t
= d
_
11
t
12
t
12
t
22
t
_
=
__
11
12
21
22
__
11
21
12
22
_
+
_
M
11
M
12
M
21
M
22
__
11
t
12
t
12
t
22
t
_
+
_
11
t
12
t
12
t
22
t
__
M
11
M
21
M
12
M
22
__
dt
+
_
11
t
12
t
12
t
22
t
_
1/2
_
dW
11
t
dW
12
t
dW
21
t
dW
22
t
__
Q
11
Q
12
Q
21
Q
22
_
+
_
Q
11
Q
21
Q
12
Q
22
__
dW
11
t
dW
21
t
dW
12
t
dW
22
t
__
11
t
12
t
12
t
22
t
_
1/2
.
Let be
_
11
t
12
t
12
t
22
t
_
:=
_
11
t
12
t
12
t
22
t
_
1/2
,
so that
2
t
=
t
=
_
(
11
t
)
2
+ (
12
t
)
2
11
t
12
t
+
12
t
22
t
11
t
12
t
+
12
t
22
t
(
12
t
)
2
+ (
22
t
)
2
_
. (61)
We rst compute the diusion coecient since it follows easily from the compu
tations in the Appendix B of Da Fonseca et al. (2005). From the dynamics of the
30
Wishart process we have:
d
11
t
= (.)dt + 2
11
t
_
Q
11
dW
11
t
+Q
21
dW
12
t
_
+ 2
12
t
_
Q
11
dW
21
t
+Q
21
dW
22
t
_
,
d
12
t
= (.)dt +
11
t
_
Q
12
dW
11
t
+Q
22
dW
12
t
_
+
12
t
_
Q
12
dW
21
t
+Q
22
dW
22
t
_
+
12
t
_
Q
11
dW
11
t
+Q
21
dW
12
t
_
+
22
t
_
Q
11
dW
21
t
+Q
21
dW
22
t
_
,
d
22
t
= (.)dt + 2
12
t
_
Q
12
dW
11
t
+Q
22
dW
12
t
_
+ 2
22
t
_
Q
12
dW
21
t
+Q
22
dW
22
t
_
,
while the covariations are given by:
d
11
,
11
)
t
= 4
11
t
(Q
2
11
+Q
2
21
)dt,
d
12
,
12
)
t
=
_
11
t
_
Q
2
12
+Q
2
22
_
+ 2
12
t
(Q
11
Q
12
+Q
21
Q
22
) +
22
t
(Q
2
11
+Q
2
21
)
_
dt,
d
22
,
22
)
t
= 4
22
t
(Q
2
12
+Q
2
22
)dt,
d
11
,
12
)
t
=
_
2
11
t
(Q
11
Q
12
+Q
21
Q
22
) + 2
12
t
_
Q
2
11
+Q
2
21
__
dt,
d
11
,
22
)
t
= 4
12
t
(Q
11
Q
12
+Q
21
Q
22
) dt,
d
12
,
22
)
t
= 2
_
12
t
_
Q
2
12
+Q
2
22
_
+
22
t
(Q
11
Q
12
+Q
21
Q
22
)
_
dt.
Now we dierentiate both sides of the equality (
12
t
)
2
=
(
12
t
)
2
11
t
22
t
, thus obtaining
2
12
t
d
12
t
=
2
12
t
11
t
22
t
d
12
t
+
_
12
t
_
2
_
1
22
t
d
_
1
11
t
_
+
1
11
t
d
_
1
22
t
__
+ (.)dt,
so that
d
12
t
=
1
_
11
t
22
t
_
d
12
t
12
t
2
11
t
d
11
t
12
t
2
22
t
d
22
t
_
+ (.)dt.
By using the covariations among the Wishart elements we have
d
12
_
t
=
1
11
t
22
t
_
11
t
_
Q
2
12
+Q
2
22
_
+ 2
12
t
(Q
11
Q
12
+Q
21
Q
22
) +
22
t
_
Q
2
11
+Q
2
21
_
+
_
12
t
_
2
_
Q
2
11
+Q
2
21
11
t
+
Q
2
12
+Q
2
22
22
t
+ 2
12
t
11
t
22
t
(Q
11
Q
12
+Q
21
Q
22
)
_
2
12
t
11
t
_
11
t
(Q
11
Q
12
+Q
21
Q
22
) +
12
t
_
Q
2
11
+Q
2
21
__
2
12
t
22
t
_
12
t
_
Q
2
12
+Q
2
22
_
+
22
t
(Q
11
Q
12
+Q
21
Q
22
)
_
_
dt,
which leads to:
d
12
_
t
=
_
1
_
12
t
_
2
_
_
Q
2
12
+Q
2
22
22
t
+
Q
2
11
+Q
2
21
11
t
2
12
t
(Q
11
Q
12
+Q
21
Q
22
)
_
11
t
22
t
_
dt.
31
Now let us compute the drift of the process
12
t
.
We dierentiate both sides of the equality
12
t
=
12
t
11
t
22
t
and we consider the
nite variation terms:
d
12
t
=
1
_
11
t
22
t
d
12
t
+
12
t
d
_
1
_
11
t
22
t
_
+d
_
12
,
1
11
22
_
t
=
1
_
11
t
22
t
_
11
21
+
12
22
+M
21
11
t
+M
12
22
t
+ (M
11
+M
22
)
12
t
_
dt
+
12
t
_
_
1
_
22
t
_
_
1
2
_
(
11
t
)
3
_
_
_
2
11
+
2
12
+ 2M
11
11
t
+ 2M
12
12
t
_
+
1
_
11
t
_
_
1
2
_
(
22
t
)
3
_
_
_
2
21
+
2
22
+ 2M
21
12
t
+ 2M
22
22
t
_
+
3
8
_
11
t
22
t
(
11
t
)
2
d
11
_
t
+
3
8
_
11
t
22
t
(
22
t
)
2
d
22
_
t
+
1
4
_
(
11
t
22
t
)
3
d
11
,
22
_
t
_
_
dt +
1
_
22
t
_
_
1
2
_
(
11
t
)
3
_
_
d
11
,
12
_
t
+
1
_
11
t
_
_
1
2
_
(
22
t
)
3
_
_
d
12
,
22
_
t
+ Diusions.
Now we use the formulas of the covariations of the Wishart elements and we
arrive to an expression which can be written as follows:
d
12
t
=
_
A
t
_
12
t
_
2
+B
t
12
t
+C
t
_
dt + Diusions,
where
9
:
A
t
=
1
_
11
t
22
t
(Q
11
Q
12
+Q
21
Q
22
)
22
t
11
t
M
12
11
t
22
t
M
21
B
t
=
2
11
+
2
12
2
11
t
2
21
+
2
22
2
22
t
+
Q
2
11
+Q
2
21
2
11
t
+
Q
2
12
+Q
2
22
2
22
t
0
C
t
=
1
_
11
t
22
t
(
11
21
+
12
22
2 (Q
11
Q
12
+Q
21
Q
22
))
+
22
t
11
t
M
12
+
11
t
22
t
M
21
.
9
Notice that the diusion term and both the expressions for B
t
and C
t
are dierent from
the ones obtained by Buraschi et al. (2006).
32
From the denition of =
12
,
11
_
t
=
1
_
11
t
22
t
_
d
11
,
12
)
t
12
t
2
11
t
d
11
,
11
)
t
12
t
2
22
t
d
22
,
22
)
t
_
= 2
11
t
22
t
_
1
_
12
t
_
2
_
(Q
11
Q
12
+Q
21
Q
22
) dt.
Using the fact that Q GL(n, R)
10
there exists a unique couple (K,
Q) O(n)
T
n
11
such that Q = K
Q. We refer to Faraut (2006) for basic results on matrix
analysis. The law of the
t
being invariant by rotation of Q we rewrite this
covariation as
d
12
,
11
_
t
= 2
11
t
22
t
_
1
_
12
t
_
2
_
Q
12
_
Q
11
+
Q
22
_
dt.
10
GL(n, R) is the linear group, the set of invertible matrices.
11
O(n) stands for the orthogonal group ie O(n) = g GL(n, R)[g
g = I
n
and T
n
is the
set of symmetric denite positive matrices.
33
R
h
o
1
S
u
p
p
o
r
t
F r e q u e n c y
0
.
2
0
.
0
0
.
2
0
.
4
0
.
6
0 5 0 1 0 0 1 5 0 2 0 0
R
h
o
2
S
u
p
p
o
r
t
F r e q u e n c y
0
.
2
0
.
0
0
.
2
0
.
4
0
.
6
0 5 0 1 0 0 2 0 0 3 0 0
S
k
e
w
f
o
r
a
s
s
e
t
1
S
u
p
p
o
r
t
F r e q u e n c y
0
.
6
0
.
4
0
.
2
0
.
0
0
.
2
0 5 0 1 0 0 1 5 0 2 0 0
S
k
e
w
f
o
r
a
s
s
e
t
2
S
u
p
p
o
r
t
F r e q u e n c y
0
.
6
0
.
4
0
.
2
0
.
0
0
.
2
0 5 0 1 0 0 1 5 0 2 0 0
F
i
g
u
r
e
1
:
E
s
t
i
m
a
t
e
d
(
t
o
p
)
a
n
d
s
k
e
w
(
b
o
t
t
o
m
)
f
r
o
m
D
C
C
e
s
t
i
m
a
t
i
o
n
o
f
t
h
e
W
A
S
C
.
R
e
a
l
g
u
r
e
s
a
r
e
i
n
r
e
d
a
n
d
D
C
C

b
a
s
e
d
e
s
t
i
m
a
t
e
d
o
n
e
s
a
r
e
i
n
b
l
u
e
.
T
h
e
g
u
r
e
d
i
s
p
l
a
y
s
t
h
e
e
s
t
i
m
a
t
e
d
p
a
r
a
m
e
t
e
r
(
t
o
p
)
a
n
d
s
k
e
w
(
b
o
t
t
o
m
)
i
n
a
t
w
o
a
s
s
e
t
s
c
a
s
e
.
W
e
r
s
t
s
i
m
u
l
a
t
e
d
a
W
A
S
C
m
o
d
e
l
i
n
w
h
i
c
h
=
(
.
5
,
.
5
)
.
W
e
t
h
e
n
e
s
t
i
m
a
t
e
d
t
h
e
D
C
C

b
a
s
e
d
t
i
m
e
v
a
r
y
i
n
g
c
o
v
a
r
i
a
n
c
e
m
a
t
r
i
c
e
s
.
F
i
n
a
l
l
y
t
h
e
s
k
e
w
i
s
e
s
t
i
m
a
t
e
d
b
y
u
s
i
n
g
e
q
u
a
t
i
o
n
(
7
)
a
n
d
(
8
)
.
B
y
d
e
n
o
t
i
n
g
t
h
e
s
k
e
w
v
e
c
t
o
r
0
=
(
0 1
,
0 2
)
,
t
h
e
v
e
c
t
o
r
i
s
e
s
t
i
m
a
t
e
d
b
y
=
_
Q
1
d
i
a
g
_
Q
Q
_
0
.
T
h
e
t
o
p
g
u
r
e
s
p
r
e
s
e
n
t
i
n
r
e
d
t
h
e
v
a
l
u
e
f
o
r
c
h
o
s
e
n
f
o
r
t
h
e
s
i
m
u
l
a
t
i
o
n
a
n
d
i
n
b
l
u
e
t
h
e
h
i
s
t
o
g
r
a
m
o
f
t
h
e
e
s
t
i
m
a
t
i
o
n
o
b
t
a
i
n
e
d
.
T
h
e
b
e
l
o
w
g
u
r
e
s
p
r
e
s
e
n
t
i
n
r
e
d
t
h
e
e
m
p
i
r
i
c
a
l
d
i
s
t
r
i
b
u
t
i
o
n
o
f
t
h
e
s
k
e
w
e
s
t
i
m
a
t
e
d
d
i
r
e
c
t
l
y
o
n
t
h
e
s
i
m
u
l
a
t
e
d
W
A
S
C
s
a
m
p
l
e
s
a
n
d
i
n
b
l
u
e
t
h
e
e
m
p
i
r
i
c
a
l
d
i
s
t
r
i
b
u
t
i
o
n
o
f
t
h
e
D
C
C

b
a
s
e
d
s
k
e
w
e
s
t
i
m
a
t
o
r
.
34
0 200 400 600 800 1000
0
.
5
1
.
0
1
.
5
2
.
0
2
.
5
Index
V
o
l
a
t
i
l
i
t
y
Time Varying Volatilities
0 200 400 600 800 1000
0
.
0
0
.
4
0
.
8
Index
C
o
r
r
e
l
a
t
i
o
n
Time Varying Correlation
Figure 2: Time varying (simulated) volatilities (top) and correlations
(bottom).
This gure displays simulated volatilities and correlation in the two dimen
sional case (n = 2). The simulation has been produced using the parameters
used in the Monte Carlo experiments. Given
t
the dynamic covariance ma
trix, the volatilities are
_
11
t
,
_
22
t
. The correlation is obtained by computing
12
t
/
_
_
11
t
+
_
22
t
_
.
35
w
300
200
100
0
100
200
300
w
300
200
100
0
100
200
300
O
b
j
e
c
t
i
v
e
0.02
0.04
0.06
0.08
0.10
Figure 3: Integrand of the CGMM estimation criterion.
The gure displays the characteristic function with the integrated volatility pre
sented in equation (29). The parameters used for to compute this characteristic
function are those used in the Monte Carlo experiments.
36
Daily Weekly
Number of obs. 500 1500 500 1500
M
11
Bias 0.0133 0.0382 0.227 0.0706
RMSE 1.5126 1.5213 1.529 1.6011
M
12
Bias 0.04 0.014 0.048 0.036
RMSE 1.0136 1.0341 1.098 1.0896
M
21
Bias 0.056 0.04 0.085 0.036
RMSE 0.981 1.0205 1.104 1.1101
M
22
Bias 0.1667 0.1784 0.070 0.0657
RMSE 1.5378 1.5502 1.511 1.5778
Q
11
Bias 0.1077 0.1069 0.132 0.1166
RMSE 0.088 0.0872 0.476 0.3568
Q
12
Bias 0.0309 0.0287 0.022 0.0208
RMSE 0.0664 0.0675 0.554 0.4467
Q
21
Bias 3E04 6E04 0.059 0.008
RMSE 0.0876 0.0862 0.553 0.371
Q
22
Bias 0.0792 0.0772 0.057 0.0574
RMSE 0.0678 0.0661 0.455 0.4424
Bias 0.0632 0.066 0.044 0.107
RMSE 4.335 4.3472 4.546 4.4364
1
Bias 0.0055 0.007 0.035 0.0268
RMSE 0.4617 0.4913 0.641 0.8031
2
Bias 0.004 0.0265 0.063 0.0445
RMSE 0.5231 0.4734 0.653 0.6405
Table 1: Results of the Monte Carlo experiments
This table displays the results for the Monte Carlo simulations performed using the following
parameters:
0
=
_
0.0225 0.0054
0.0054 0.0144
_
.M =
_
5 3
3 5
_
, =
_
0.3 0.4
, (62)
Q =
_
0.1133137 0.03335871
0.0000000 0.07954368
_
, = 15. (63)
Two dierent types of simulations are presented: one of the sample includes 500 daily observa
tions and a second one uses 1500 daily observations, as in Carrasco et al. (2007).
SP500 FTSE DAX CAC 40
Min. :0.128129 Min. :0.141420 Min. :0.197775 Min. :0.149149
1st Qu.:0.009940 1st Qu.:0.010858 1st Qu.:0.014283 1st Qu.:0.015173
Median : 0.002491 Median : 0.002101 Median : 0.003727 Median : 0.002117
Mean : 0.001535 Mean : 0.001014 Mean : 0.001523 Mean : 0.001095
3rd Qu.: 0.013506 3rd Qu.: 0.013424 3rd Qu.: 0.019416 3rd Qu.: 0.018025
Max. : 0.123746 Max. : 0.135879 Max. : 0.171546 Max. : 0.166252
Table 2: Descriptive statistics for the real dataset.
The table summarizes the descriptive statistics for the available dataset. This dataset is made
of the SP500, FTSE, DAX and CAC time series, on a weekly sampling frequency. The dataset
starts on January 2nd 1990 and ends on June 30th 2007.
37
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
M
1
1

3
.
4
4

2
.
9
4

2
.
6
5

3
.
2
2

2
.
6
2

2
.
6
1
S
t
d
.
D
e
v
.
(
0
.
2
3
9
)
(
0
.
2
1
7
)
(
0
.
3
0
9
)
(
0
.
1
5
)
(
0
.
1
5
8
)
(
0
.
1
8
1
)
M
1
2
1
.
1
7
0
.
3
3
0
.
7
4
1
.
5
1
0
.
0
4
0
.
5
S
t
d
.
D
e
v
.
(
0
.
2
1
1
)
(
0
.
2
6
7
)
(
0
.
5
1
9
)
(
0
.
2
6
6
)
(
0
.
3
5
8
)
(
0
.
2
3
6
)
M
2
1
0
.
2
2
0
.
9
6
1
.
7
9
1
.
5
1
0
.
7
4
0
.
7
6
S
t
d
.
D
e
v
.
(
0
.
4
6
9
)
(
0
.
3
7
2
)
(
0
.
4
9
1
)
(
0
.
1
6
8
)
(
0
.
1
5
3
)
(
0
.
1
6
2
)
M
2
2

2
.
8
7

3
.
4
1

3
.
0
4

3
.
4
9

3
.
5
3

2
.
4
8
S
t
d
.
D
e
v
.
(
0
.
2
1
)
(
0
.
2
6
1
)
(
0
.
3
8
4
)
(
0
.
1
9
3
)
(
0
.
2
1
5
)
(
0
.
1
6
5
)
Q
1
1

0
.
0
5

0
.
0
7

0
.
0
1
0
.
0
6
0
.
0
6
0
.
1
2
S
t
d
.
D
e
v
.
(
0
.
0
1
)
(
0
.
0
0
5
)
(
0
.
0
0
8
)
(
0
.
0
0
4
)
(
0
.
0
0
5
)
(
0
.
0
0
3
)
Q
1
2

0
.
1
0
.
0
1

0
.
0
8

0
.
0
4

0
.
0
4
0
.
0
4
S
t
d
.
D
e
v
.
(
0
.
0
1
2
)
(
0
.
0
0
7
)
(
0
.
0
0
7
)
(
0
.
0
0
4
)
(
0
.
0
0
4
)
(
0
.
0
0
3
)
Q
2
1
0
.
0
7
0
.
0
2

0
.
0
8

0
.
0
8
0
.
0
9
0
.
0
1
S
t
d
.
D
e
v
.
(
0
.
0
0
7
)
(
0
.
0
0
5
)
(
0
.
0
0
7
)
(
0
.
0
0
7
)
(
0
.
0
0
9
)
(
0
.
0
0
4
)
Q
2
2
0
.
0
7
0
.
1
2

0
.
0
3

0
.
1
0
.
0
8
0
.
0
6
S
t
d
.
D
e
v
.
(
0
.
0
1
4
)
(
0
.
0
0
5
)
(
0
.
0
1
)
(
0
.
0
0
4
)
(
0
.
0
0
5
)
(
0
.
0
0
3
)
1
0
.
6
5
1
2
.
2
5
1
2
.
7
1
1
.
3
4
1
0
.
7
4
1
0
.
8
7
S
t
d
.
D
e
v
.
(
0
.
7
3
3
)
(
0
.
7
7
6
)
(
1
.
2
3
6
)
(
0
.
6
6
4
)
(
0
.
7
3
8
)
(
0
.
6
3
8
)
1
0
.
3
0
.
6
0
.
4
0
.
1
6

0
.
2

0
.
3
S
t
d
.
D
e
v
.
(
0
.
0
6
7
)
(
0
.
1
0
3
)
(
0
.
0
9
8
)
(
0
.
0
7
6
)
(
0
.
0
8
1
)
(
0
.
0
6
9
)
2

0
.
2

0
.
4
0
.
2
8
0
.
4

0
.
2
6

0
.
8
1
S
t
d
.
D
e
v
.
(
0
.
0
7
1
)
(
0
.
0
6
3
)
(
0
.
0
8
1
)
(
0
.
0
6
)
(
0
.
0
7
6
)
(
0
.
1
1
)
T
a
b
l
e
3
:
H
i
s
t
o
r
i
c
a
l
p
a
r
a
m
e
t
e
r
e
s
t
i
m
a
t
e
s
f
o
r
t
h
e
W
A
S
C
m
o
d
e
l
u
s
i
n
g
t
h
e
d
a
i
l
y
d
a
t
a
s
e
t
.
T
h
e
t
a
b
l
e
p
r
e
s
e
n
t
s
t
h
e
p
a
r
a
m
e
t
e
r
s
e
s
t
i
m
a
t
e
d
u
s
i
n
g
t
h
e
C

G
M
M
m
e
t
h
o
d
d
e
v
e
l
o
p
e
d
i
n
C
a
r
r
a
s
c
o
e
t
a
l
.
(
2
0
0
7
)
,
u
s
i
n
g
a
s
i
m
p
l
e
i
n
d
e
x
p
r
o
c
e
d
u
r
e
.
T
h
e
i
n
t
e
g
r
a
t
i
o
n
g
r
i
d
i
s
=
[
3
0
0
;
3
0
0
]
.
T
h
e
d
a
t
a
s
e
t
s
a
r
e
m
a
d
e
o
f
d
a
i
l
y
o
b
s
e
r
v
a
t
i
o
n
s
,
f
r
o
m
0
1
/
0
1
/
1
9
9
0
u
n
t
i
l
0
9
/
1
2
/
2
0
0
7
,
u
s
i
n
g
a
c
t
i
v
e
d
a
y
s
c
l
o
s
i
n
g
p
r
i
c
e
s
.
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
s
a
r
e
g
i
v
e
n
b
e
t
w
e
e
n
b
r
a
c
k
e
t
s
.
38
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
M
1
1

2
.
8
3

2
.
6
7

3
.
4
3

3
.
1
4

2
.
8
9

3
.
4
2
S
t
d
.
D
e
v
.
(
0
.
6
7
1
)
(
1
.
0
7
7
)
(
0
.
6
7
1
)
(
0
.
3
2
6
)
(
0
.
2
7
9
)
(
0
.
2
6
7
)
M
1
2
1
.
8
7
1
.
2
8
1
.
1
3
0
.
4
4
0
.
8
3
0
.
0
1
S
t
d
.
D
e
v
.
(
0
.
3
7
6
)
(
0
.
5
6
6
)
(
0
.
5
3
8
)
(
0
.
3
4
3
)
(
0
.
5
6
4
)
(
0
.
5
2
7
)
M
2
1
0
.
4
2
1
.
5
4
0
.
8
2
0
.
6
3
0
.
8
9
1
.
0
9
S
t
d
.
D
e
v
.
(
1
.
0
0
2
)
(
1
.
6
6
1
)
(
0
.
7
2
)
(
0
.
3
8
)
(
0
.
2
6
4
)
(
0
.
2
6
)
M
2
2

3
.
2
9

3
.
7
9

3
.
2
1

3
.
9

1
.
2
7

3
.
7
6
S
t
d
.
D
e
v
.
(
0
.
3
0
5
)
(
0
.
3
9
3
)
(
0
.
2
9
1
)
(
0
.
2
6
4
)
(
0
.
3
2
6
)
(
0
.
3
1
2
)
Q
1
1
0
.
0
4

0
.
0
5
0
.
0
7
0
.
0
1
0
.
0
4

0
.
1
2
S
t
d
.
D
e
v
.
(
0
.
0
0
3
)
(
0
.
0
1
2
)
(
0
.
0
0
6
)
(
0
.
0
0
3
)
(
0
.
0
1
1
)
(
0
.
0
0
6
)
Q
1
2
0
.
0
5
0
.
0
2
0
.
0
8

0
.
1
5
0
.
0
6

0
.
0
5
S
t
d
.
D
e
v
.
(
0
.
0
0
5
)
(
0
.
0
1
4
)
(
0
.
0
0
5
)
(
0
.
0
0
2
)
(
0
.
0
0
7
)
(
0
.
0
0
5
)
Q
2
1
0
0
.
0
3

0
.
0
2

0
.
1
1
0
.
0
9

0
.
0
2
S
t
d
.
D
e
v
.
(
0
.
0
0
5
)
(
0
.
0
0
4
)
(
0
.
0
0
4
)
(
0
.
0
0
7
)
(
0
.
0
0
6
)
(
0
.
0
0
6
)
Q
2
2

0
.
1
2
0
.
1
3

0
.
0
5

0
.
1
0
.
0
1

0
.
0
7
S
t
d
.
D
e
v
.
(
0
.
0
0
7
)
(
0
.
0
0
6
)
(
0
.
0
0
4
)
(
0
.
0
0
8
)
(
0
.
0
0
5
)
(
0
.
0
0
4
)
1
2
.
4
4
1
3
.
5
8
1
3
.
1
7
1
4
.
8
5
1
1
.
5
6
1
4
.
3
7
S
t
d
.
D
e
v
.
(
1
.
2
0
5
)
(
1
.
3
7
6
)
(
1
.
9
2
9
)
(
0
.
3
9
2
)
(
1
.
8
4
7
)
(
1
.
2
9
)
1

0
.
7
5
0
.
0
1

0
.
4
0
.
3
9

0
.
2
3
0
.
3
6
S
t
d
.
D
e
v
.
(
0
.
0
6
7
)
(
0
.
0
4
1
)
(
0
.
0
3
1
)
(
0
.
0
3
8
)
(
0
.
0
7
8
)
(
0
.
0
6
2
)
2
0
.
3
5

0
.
6
2
0
.
6
4
0
.
2
6

0
.
4
9
0
.
1
1
S
t
d
.
D
e
v
.
(
0
.
0
3
5
)
(
0
.
0
3
9
)
(
0
.
0
5
8
)
(
0
.
0
7
6
)
(
0
.
0
6
5
)
(
0
.
0
5
4
)
T
a
b
l
e
4
:
H
i
s
t
o
r
i
c
a
l
p
a
r
a
m
e
t
e
r
e
s
t
i
m
a
t
e
s
f
o
r
t
h
e
W
A
S
C
m
o
d
e
l
u
s
i
n
g
t
h
e
w
e
e
k
l
y
d
a
t
a
s
e
t
.
T
h
e
t
a
b
l
e
p
r
e
s
e
n
t
s
t
h
e
p
a
r
a
m
e
t
e
r
s
e
s
t
i
m
a
t
e
d
u
s
i
n
g
t
h
e
C

G
M
M
m
e
t
h
o
d
d
e
v
e
l
o
p
e
d
i
n
C
a
r
r
a
s
c
o
e
t
a
l
.
(
2
0
0
7
)
,
u
s
i
n
g
a
s
i
m
p
l
e
i
n
d
e
x
p
r
o
c
e
d
u
r
e
.
T
h
e
i
n
t
e
g
r
a
t
i
o
n
g
r
i
d
i
s
=
[
3
0
0
;
3
0
0
]
.
T
h
e
d
a
t
a
s
e
t
s
a
r
e
m
a
d
e
o
f
w
e
e
k
l
y
o
b
s
e
r
v
a
t
i
o
n
s
,
f
r
o
m
0
1
/
0
1
/
1
9
9
0
u
n
t
i
l
0
9
/
1
2
/
2
0
0
7
,
u
s
i
n
g
a
c
t
i
v
e
d
a
y
s
c
l
o
s
i
n
g
p
r
i
c
e
s
.
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
s
a
r
e
g
i
v
e
n
b
e
t
w
e
e
n
b
r
a
c
k
e
t
s
.
39
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
D
a
i
l
y
V
o
l
v
o
l
A
s
s
e
t
1
0
.
1
7
2
0
.
1
4
6
0
.
1
6
1
0
.
2
0
.
2
1
6
0
.
2
4
1
V
o
l
v
o
l
A
s
s
e
t
2
0
.
0
8
0
.
1
1
7
0
.
1
4
6
0
.
2
2
1
0
.
1
9
7
0
.
2
4
3
W
e
e
k
l
y
V
o
l
v
o
l
A
s
s
e
t
1
0
.
2
4
4
0
.
2
4
1
0
.
1
7
1
0
.
2
1
5
0
.
1
7
9
0
.
1
4
4
V
o
l
v
o
l
A
s
s
e
t
2
0
.
2
6
0
.
2
6
3
0
.
1
8
9
0
.
3
6
1
0
.
1
2
2
0
.
1
7
2
T
a
b
l
e
5
:
V
o
l
a
t
i
l
i
t
y
o
f
v
o
l
a
t
i
l
i
t
y
e
s
t
i
m
a
t
e
d
f
r
o
m
t
h
e
d
i
e
r
e
n
t
d
a
t
a
s
e
t
s
,
w
i
t
h
d
i
e
r
e
n
t
s
a
m
p
l
e
f
r
e
q
u
e
n
c
y
.
I
n
t
h
e
t
a
b
l
e
,
t
h
e
r
s
t
l
i
n
e
c
o
r
r
e
s
p
o
n
d
s
t
o
t
h
e
v
o
l
a
t
i
l
i
t
y
o
f
v
o
l
a
t
i
l
i
t
y
o
f
t
h
e
r
s
t
a
s
s
e
t
.
F
o
r
e
x
a
m
p
l
e
,
f
o
r
t
h
e
c
o
l
u
m
n
S
P
5
0
0
/
C
A
C
,
t
h
e
r
s
t
l
i
n
e
c
o
r
r
e
s
p
o
n
d
s
t
o
t
h
e
v
o
l
a
t
i
l
i
t
y
o
f
v
o
l
a
t
i
l
i
t
y
o
f
t
h
e
S
P
5
0
0
.
T
h
e
s
e
c
o
n
d
l
i
n
e
d
i
s
p
l
a
y
s
t
h
e
o
n
e
o
b
t
a
i
n
e
d
f
o
r
t
h
e
C
A
C
.
T
h
e
v
o
l
a
t
i
l
i
t
y
o
f
v
o
l
a
t
i
l
i
t
y
i
s
o
b
t
a
i
n
e
d
w
h
e
n
c
o
m
p
u
t
i
n
g
:
V
o
l
/
v
o
l
A
s
s
e
t
1
=
2
_
Q
2 1
1
+
Q
2 2
1
V
o
l
/
v
o
l
A
s
s
e
t
2
=
2
_
Q
2 1
2
+
Q
2 2
2
.
40
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
D
a
i
l
y
S
k
e
w
A
s
s
e
t
1

0
.
3
3
7

0
.
6
8
7

0
.
3
2
7

0
.
2
2
4

0
.
3
2
7

0
.
3
6
6
S
k
e
w
A
s
s
e
t
2

0
.
7
5

0
.
3
2
8

0
.
5
6

0
.
2
1
9

0
.
5
4
1

0
.
3
7
3
W
e
e
k
l
y
S
k
e
w
A
s
s
e
t
1

0
.
3
6

0
.
3
4
9

0
.
4
6
8

0
.
4
3
1

0
.
1
4
3

0
.
8
4
S
k
e
w
A
s
s
e
t
2

0
.
6
1
2

0
.
6
1
1

0
.
6
7
8

0
.
4
6
6

0
.
3
0
7

0
.
2
9
9
T
a
b
l
e
6
:
S
k
e
w
e
s
t
i
m
a
t
e
d
f
r
o
m
t
h
e
d
i
e
r
e
n
t
d
a
t
a
s
e
t
s
,
w
i
t
h
d
i
e
r
e
n
t
s
a
m
p
l
e
f
r
e
q
u
e
n
c
y
.
I
n
t
h
e
t
a
b
l
e
,
t
h
e
r
s
t
l
i
n
e
c
o
r
r
e
s
p
o
n
d
s
t
o
t
h
e
s
k
e
w
a
s
s
o
c
i
a
t
e
d
t
o
t
h
e
r
s
t
a
s
s
e
t
.
F
o
r
e
x
a
m
p
l
e
,
f
o
r
t
h
e
c
o
l
u
m
n
S
P
5
0
0
/
C
A
C
,
t
h
e
r
s
t
l
i
n
e
c
o
r
r
e
s
p
o
n
d
s
t
o
t
h
e
s
k
e
w
o
f
t
h
e
S
P
5
0
0
.
T
h
e
s
e
c
o
n
d
l
i
n
e
d
i
s
p
l
a
y
s
t
h
e
o
n
e
o
b
t
a
i
n
e
d
f
o
r
t
h
e
C
A
C
.
T
h
e
s
k
e
w
i
s
o
b
t
a
i
n
e
d
w
h
e
n
c
o
m
p
u
t
i
n
g
:
C
o
r
r
_
d
Y
1
t
,
d
1
1
t
_
=
Q
1
1
1
+
Q
2
1
2
_
Q
2 1
1
+
Q
2 2
1
C
o
r
r
_
d
Y
2
t
,
d
2
2
t
_
=
Q
1
2
1
+
Q
2
2
2
_
Q
2 1
2
+
Q
2 2
2
.
41
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
D
a
i
l
y
1
1
0
.
1
5
8
0
.
1
5
8
0
.
1
5
8
0
.
2
2
1
0
.
2
2
1
0
.
2
0
6
2
2
0
.
2
2
1
0
.
2
0
6
0
.
1
5
8
0
.
2
0
6
0
.
1
5
8
0
.
1
5
8
1
2
0
.
4
4
9
0
.
4
1
7
0
.
4
0
5
0
.
7
6
6
0
.
6
6
9
0
.
7
6
1
W
e
e
k
l
y
1
1
0
.
1
8
0
.
1
8
0
.
1
8
0
.
2
6
0
.
2
6
0
.
2
3
8
2
2
0
.
2
6
0
.
2
3
8
0
.
1
8
0
.
2
3
8
0
.
1
8
0
.
1
8
1
2
0
.
6
9
4
0
.
6
7
6
0
.
6
7
4
0
.
8
3
1
0
.
7
3
9
0
.
7
8
1
T
a
b
l
e
7
:
L
o
n
g
t
e
r
m
c
o
v
a
r
i
a
n
c
e
m
a
t
r
i
x
i
m
p
l
i
e
d
b
y
t
h
e
e
s
t
i
m
a
t
e
s
.
T
h
e
l
o
n
g
t
e
r
m
c
o
v
a
r
i
a
n
c
e
m
a
t
r
i
x
i
s
e
s
t
i
m
a
t
e
d
t
h
r
o
u
g
h
t
h
e
u
s
u
a
l
m
o
m
e
n
t
e
s
t
i
m
a
t
o
r
s
.
I
t
s
d
i
a
g
o
n
a
l
e
l
e
m
e
n
t
s
a
r
e
1
1
t
a
n
d
2
2
t
.
(
1
2
t
)
i
s
t
h
e
l
o
n
g
t
e
r
m
c
o
r
r
e
l
a
t
i
o
n
c
o
m
p
u
t
e
d
f
r
o
m
t
h
e
e
s
t
i
m
a
t
e
d
l
o
n
g
t
e
r
m
c
o
v
a
r
i
a
n
c
e
m
a
t
r
i
x
.
42
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
A
s
s
e
t
1
A
s
s
e
t
2
A
s
s
e
t
1
A
s
s
e
t
2
A
s
s
e
t
1
A
s
s
e
t
2
A
s
s
e
t
1
A
s
s
e
t
2
A
s
s
e
t
1
A
s
s
e
t
2
A
s
s
e
t
1
A
s
s
e
t
2
D
a
i
l
y
1
6
.
8
8
5
.
7
4
5
.
8
8
6
.
8
2
5
.
3
6
.
0
8
6
.
4
4
6
.
9
8
5
.
2
4
7
.
0
6
5
.
2
2
4
.
9
6
2

1
.
4
9
3

0
.
7
4

0
.
3
6
3

0
.
2
0
8

0
.
6
1
2

0
.
5
8
6

2
.
1
4
9

2
.
4
8
9

0
.
0
3
8

0
.
0
7
4

0
.
5
8
9

0
.
9
8
3
5
.
3
8
7
5
5
.
5
1
7
6
.
6
1
2
4
.
6
8
8
5
.
4
9
4
4
.
2
9
1
4
.
4
9
1
5
.
2
0
2
6
.
9
8
6
4
.
6
3
1
3
.
9
7
7
W
e
e
k
l
y
1
5
.
6
6
6
.
5
8
5
.
3
4
7
.
5
8
2
0
.
8
6
6
.
4
2
6
.
2
8
7
.
8
5
.
7
8
2
.
5
4
6
.
8
4
7
.
5
2
2

3
.
7
3
9

1
.
8
0
4

2
.
2
8
8

1
.
3
0
8

1
.
5
0
9

1
.
5
3
6

0
.
6
7
2

0
.
7
9
6

0
.
8
4
4

1
.
7
8
1

0
.
0
1
2

0
.
0
2
1
1
.
9
2
1
4
.
7
7
6
3
.
0
5
2
6
.
2
7
2
1
9
.
3
5
1
4
.
8
8
4
5
.
6
0
8
7
.
0
0
4
4
.
9
3
6
0
.
7
5
9
6
.
8
2
8
7
.
4
9
9
T
a
b
l
e
8
:
M
e
a
n
r
e
v
e
r
t
i
n
g
p
a
r
a
m
e
t
e
r
a
s
s
o
c
i
a
t
e
d
t
o
t
.
T
h
i
s
t
a
b
l
e
d
i
s
p
l
a
y
s
t
h
e
m
e
a
n
r
e
v
e
r
t
i
n
g
p
a
r
a
m
e
t
e
r
a
s
s
o
c
i
a
t
e
d
t
o
t
h
e
v
o
l
a
t
i
l
i
t
i
e
s
.
F
o
r
t
h
e
r
s
t
a
s
s
e
t
,
w
e
h
a
v
e
:
d
1
1
t
=
.
.
.
+
1
1
t
_ _ _ _ _ _ _ _ _ _
2
M
1
1
.
1
+
2
M
1
2
2
2
t
1
1
t
(
1
2
t
)
.
2
.
_ _ _ _ _ _ _ _ _ _
d
t
+
.
.
.
1
i
s
t
h
e
i
d
i
o
s
y
n
c
r
a
t
i
c
p
a
r
t
o
f
t
h
e
m
e
a
n
r
e
v
e
r
t
i
n
g
p
a
r
a
m
e
t
e
r
.
2
i
s
t
h
e
c
r
o
s
s

c
o
m
p
o
n
e
n
t
i
n
t
h
e
p
a
r
a
m
e
t
e
r
a
n
d
i
s
t
h
e
s
t
a
n
d
a
r
d
H
e
s
t
o
n

l
i
k
e
m
e
a
n
r
e
v
e
r
t
i
n
g
p
a
r
a
m
e
t
e
r
.
(
1
2
)
=
C
o
r
r
(
d
l
o
g
S
1
,
d
l
o
g
S
2
)
i
s
t
h
e
l
o
n
g
t
e
r
m
c
o
r
r
e
l
a
t
i
o
n
b
e
t
w
e
e
n
t
h
e
a
s
s
e
t
l
o
g
r
e
t
u
r
n
s
.
T
h
e
d
r
i
f
t
i
s
e
v
a
l
u
a
t
e
d
a
t
i
t
s
s
t
a
t
i
o
n
a
r
y
v
a
l
u
e
,
u
s
i
n
g
t
h
e
v
a
l
u
e
s
p
r
e
s
e
n
t
e
d
i
n
t
a
b
l
e
7
.
43
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
D
a
i
l
y
A
(
t
)

1
.
5
2
2

1
.
1
0
8

2
.
3
7
6

2
.
9
0
2

0
.
9
1
8

1
.
1
9
9
B
(
t
)

3
.
0
2
2

3
.
2
4
6

3
.
3
4
5

2
.
5
4
4

2
.
7
7
9

2
.
7
8
2
C
(
t
)
4
.
3
7
1
1
.
7
2
2
3
.
9
2
9
4
.
2
1
2
.
2
8
3
2
.
8
7
4
W
e
e
k
l
y
A
(
t
)

2
.
9
2
2

2
.
7
6

1
.
0
5
6

0
.
8
6
8

1
.
7
6
1

1
.
2
1
B
(
t
)

2
.
4
8
5

3
.
7
2
3

3
.
8
9

7
.
5
4
5

1
.
9
8
6

4
.
7
6
1
C
(
t
)
3
.
6
2
9
3
.
9
8
8
4
.
6
6
4
3
.
9
4
5
2
.
8
4
9
4
.
5
7
1
T
a
b
l
e
9
:
M
e
a
n
r
e
v
e
r
t
i
n
g
p
a
r
a
m
e
t
e
r
s
a
s
s
o
c
i
a
t
e
d
t
o
1
2
t
.
T
h
i
s
t
a
b
l
e
d
i
s
p
l
a
y
s
t
h
e
d
r
i
f
t
p
a
r
t
o
f
t
h
e
c
o
r
r
e
l
a
t
i
o
n
c
o
e
c
i
e
n
t
(
1
2
t
)
i
n
t
h
e
t
w
o

a
s
s
e
t
c
a
s
e
,
f
o
r
t
h
e
v
a
l
u
e
e
s
t
i
m
a
t
e
d
i
n
t
h
e
p
r
e
v
i
o
u
s
t
a
b
l
e
s
.
T
h
e
d
r
i
f
t
i
s
g
i
v
e
n
b
y
:
d
1
2
t
=
(
A
t
(
1
2
t
)
2
+
B
t
(
1
2
t
)
+
C
t
)
d
t
+
d
i
u
s
i
o
n
p
a
r
t
.
T
h
e
c
o
m
p
u
t
a
t
i
o
n
o
f
t
h
i
s
d
r
i
f
t
i
s
p
r
e
s
e
n
t
e
d
i
n
t
h
e
a
p
p
e
n
d
i
c
e
s
.
T
h
e
d
r
i
f
t
i
s
c
o
m
p
u
t
e
d
a
t
i
t
s
s
t
a
t
i
o
n
a
r
y
v
a
l
u
e
,
u
s
i
n
g
t
h
e
v
a
l
u
e
s
p
r
e
s
e
n
t
e
d
i
n
t
a
b
l
e
7
.
44
0
.
8
0
.
9
1
.
0
1
.
1
1 . 0 0 . 0 0 . 5 1 . 0 1 . 5 2 . 0
S
P
5
0
0
D
A
X
r
h
o
d r h o
0
.
4
0
0
.
4
5
0
.
5
0
0
.
5
5
0 . 4 0 . 2 0 . 0 0 . 2 0 . 4
S
P
5
0
0
C
A
C
r
h
o
d r h o
0
.
6
0
0
.
6
5
0
.
7
0
0
.
7
5
0
.
8
0
0
.
8
5
0
.
9
0
1 . 0 0 . 0 0 . 5 1 . 0 1 . 5 2 . 0
S
P
5
0
0
F
T
S
E
r
h
o
d r h o
0
.
7
0
0
.
8
0
0
.
9
0
1
.
0
0
1 0 1 2
D
A
X
C
A
C
r
h
o
d r h o
0
.
5
5
0
.
6
0
0
.
6
5
0
.
7
0
0
.
7
5
0
.
8
0
0 . 6 0 . 2 0 . 2 0 . 6
D
A
X
F
T
S
E
r
h
o
d r h o
0
.
6
5
0
.
7
0
0
.
7
5
0
.
8
0
0
.
8
5
0
.
9
0
0 . 5 0 . 0 0 . 5 1 . 0
C
A
C
F
T
S
E
r
h
o
d r h o
F
i
g
u
r
e
4
:
d
(
1
2
t
)
a
s
a
f
u
n
c
t
i
o
n
o
f
(
1
2
t
)
u
s
i
n
g
t
h
e
d
a
i
l
y
d
a
t
a
s
e
t
.
T
h
i
s
g
u
r
e
d
i
s
p
l
a
y
s
t
h
e
d
r
i
f
t
t
e
r
m
a
s
s
o
c
i
a
t
e
d
t
o
d
(
1
2
)
a
s
a
f
u
n
c
t
i
o
n
o
f
(
1
2
)
,
u
s
i
n
g
d
a
i
l
y
e
s
t
i
m
a
t
i
o
n
r
e
s
u
l
t
s
.
T
h
i
s
r
e
l
a
t
i
o
n
i
s
s
p
e
c
i
e
d
i
n
t
h
e
f
o
l
l
o
w
i
n
g
e
q
u
a
t
i
o
n
:
d
1
2
t
=
(
A
t
(
1
2
t
)
2
+
B
t
(
1
2
t
)
+
C
t
)
d
t
+
d
i
u
s
i
o
n
p
a
r
t
.
T
h
e
b
l
u
e
l
i
n
e
i
s
t
h
e
l
i
n
e
a
r
t
e
r
m
o
f
t
h
e
l
a
t
t
e
r
r
e
l
a
t
i
o
n
.
T
h
e
r
e
d
o
n
e
i
s
t
h
e
t
o
t
a
l
e
x
p
e
c
t
e
d
v
a
r
i
a
t
i
o
n
o
f
d
(
1
2
t
)
d
e
p
e
n
d
i
n
g
u
p
o
n
t
h
e
(
1
2
t
)
v
a
l
u
e
s
.
T
h
e
g
r
e
e
n
l
i
n
e
s
i
g
n
a
l
s
t
h
e
l
o
n
g
t
e
r
m
c
o
r
r
e
l
a
t
i
o
n
i
m
p
l
i
e
d
b
y
t
h
e
e
s
t
i
m
a
t
e
d
W
A
S
C
m
o
d
e
l
.
T
h
i
s
l
o
n
g
t
e
r
m
c
o
r
r
e
l
a
t
i
o
n
i
s
t
h
e
o
n
l
y
p
o
s
i
t
i
v
e
r
o
o
t
o
f
t
h
e
p
o
l
y
n
o
m
i
a
l
d
e
n
e
d
b
y
t
h
e
s
t
a
t
i
o
n
a
r
y
d
r
i
f
t
o
f
t
h
e
c
o
r
r
e
l
a
t
i
o
n
,
u
s
i
n
g
t
h
e
v
a
l
u
e
s
p
r
e
s
e
n
t
e
d
i
n
t
a
b
l
e
7
.
45
0
.
6
5
0
.
7
0
0
.
7
5
0
.
8
0
0
.
8
5
0
.
9
0
1 . 0 0 . 0 0 . 5 1 . 0 1 . 5 2 . 0
S
P
5
0
0
D
A
X
r
h
o
d r h o
0
.
6
0
0
.
6
5
0
.
7
0
0
.
7
5
0
.
8
0
0
.
8
5
1 . 0 0 . 0 0 . 5 1 . 0 1 . 5 2 . 0
S
P
5
0
0
C
A
C
r
h
o
d r h o
0
.
8
0
.
9
1
.
0
1
.
1
1 . 0 0 . 0 0 . 5 1 . 0 1 . 5
S
P
5
0
0
F
T
S
E
r
h
o
d r h o
0
.
4
0
0
.
4
5
0
.
5
0
0
.
5
5
0
.
6
0
0 . 5 0 . 0 0 . 5 1 . 0
D
A
X
C
A
C
r
h
o
d r h o
0
.
6
5
0
.
7
5
0
.
8
5
0
.
9
5
0 . 5 0 . 0 0 . 5 1 . 0 1 . 5
D
A
X
F
T
S
E
r
h
o
d r h o
0
.
6
5
0
.
7
0
0
.
7
5
0
.
8
0
0
.
8
5
0
.
9
0
0
.
9
5
1 . 0 0 . 5 0 . 0 0 . 5 1 . 0 1 . 5
C
A
C
F
T
S
E
r
h
o
d r h o
F
i
g
u
r
e
5
:
d
(
1
2
t
)
a
s
a
f
u
n
c
t
i
o
n
o
f
(
1
2
t
)
u
s
i
n
g
t
h
e
w
e
e
k
l
y
d
a
t
a
s
e
t
.
T
h
i
s
g
u
r
e
d
i
s
p
l
a
y
s
t
h
e
d
r
i
f
t
t
e
r
m
a
s
s
o
c
i
a
t
e
d
t
o
d
(
1
2
t
)
a
s
a
f
u
n
c
t
i
o
n
o
f
(
1
2
t
)
,
u
s
i
n
g
w
e
e
k
l
y
e
s
t
i
m
a
t
i
o
n
r
e
s
u
l
t
s
.
T
h
i
s
r
e
l
a
t
i
o
n
i
s
s
p
e
c
i
e
d
i
n
t
h
e
f
o
l
l
o
w
i
n
g
e
q
u
a
t
i
o
n
:
d
1
2
=
(
A
t
(
1
2
t
)
2
+
B
t
(
1
2
t
)
+
C
t
)
d
t
+
d
i
u
s
i
o
n
p
a
r
t
.
T
h
e
b
l
u
e
l
i
n
e
i
s
t
h
e
l
i
n
e
a
r
t
e
r
m
o
f
t
h
e
l
a
t
t
e
r
r
e
l
a
t
i
o
n
.
T
h
e
r
e
d
o
n
e
i
s
t
h
e
t
o
t
a
l
e
x
p
e
c
t
e
d
v
a
r
i
a
t
i
o
n
o
f
d
(
1
2
t
)
d
e
p
e
n
d
i
n
g
u
p
o
n
t
h
e
(
1
2
t
)
v
a
l
u
e
s
.
T
h
e
g
r
e
e
n
l
i
n
e
s
i
g
n
a
l
s
t
h
e
l
o
n
g
t
e
r
m
c
o
r
r
e
l
a
t
i
o
n
i
m
p
l
i
e
d
b
y
t
h
e
e
s
t
i
m
a
t
e
d
W
A
S
C
m
o
d
e
l
.
T
h
i
s
l
o
n
g
t
e
r
m
c
o
r
r
e
l
a
t
i
o
n
i
s
t
h
e
o
n
l
y
p
o
s
i
t
i
v
e
r
o
o
t
o
f
t
h
e
p
o
l
y
n
o
m
i
a
l
d
e
n
e
d
b
y
t
h
e
s
t
a
t
i
o
n
a
r
y
d
r
i
f
t
o
f
t
h
e
c
o
r
r
e
l
a
t
i
o
n
,
u
s
i
n
g
t
h
e
v
a
l
u
e
s
p
r
e
s
e
n
t
e
d
i
n
t
a
b
l
e
7
.
46
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
D
a
i
l
y
0
.
3
6
1
0
.
7
2
1
0
.
4
8
4
0
.
4
3
1
0
.
3
2
8
0
.
8
6
4
W
e
e
k
l
y
0
.
8
2
8
0
.
6
2
0
0
.
7
5
5
0
.
4
6
6
0
.
5
4
1
0
.
3
7
6
T
a
b
l
e
1
0
:
N
o
r
m
o
f
t
h
e
v
e
c
t
o
r
.
T
h
e
n
o
r
m
o
f
i
s
c
o
m
p
u
t
e
d
a
s
[
[
[
[
=
_
(
)
,
w
h
i
c
h
i
s
t
h
e
E
u
c
l
i
d
i
a
n
n
o
r
m
.
47
S
P
5
0
0
/
D
A
X
S
P
5
0
0
/
C
A
C
S
P
5
0
0
/
F
T
S
E
D
A
X
/
C
A
C
F
T
S
E
/
D
A
X
F
T
S
E
/
C
A
C
D
a
i
l
y
0
.
0
5
8
0
.
0
0
9
0
.
0
2
0
.
0
2
8
0
.
0
2
5
0
.
0
3
W
e
e
k
l
y
0
.
0
1
2
0
.
0
1
6
0
.
0
4
9
0
.
0
3
4
0
.
0
2
2
0
.
0
3
8
T
a
b
l
e
1
1
:
O
d
i
a
g
o
n
a
l
t
e
r
m
o
f
t
h
e
p
o
l
a
r
d
e
c
o
m
p
o
s
i
t
i
o
n
o
f
Q
e
s
t
i
m
a
t
e
d
f
r
o
m
t
h
e
d
i
e
r
e
n
t
d
a
t
a
s
e
t
s
,
w
i
t
h
d
i
e
r
e
n
t
s
a
m
p
l
e
f
r
e
q
u
e
n
c
y
.
U
s
i
n
g
t
h
e
f
a
c
t
t
h
a
t
Q
G
L
(
n
,
R
)
,
w
h
e
r
e
G
L
(
n
,
R
)
i
s
t
h
e
l
i
n
e
a
r
g
r
o
u
p
,
i
.
e
.
t
h
e
s
e
t
o
f
i
n
v
e
r
t
i
b
l
e
m
a
t
r
i
c
e
s
,
t
h
e
r
e
e
x
i
s
t
s
a
u
n
i
q
u
e
c
o
u
p
l
e
(
K
,
Q
)
O
(
n
)
T
n
,
w
h
e
r
e
O
(
n
)
s
t
a
n
d
s
f
o
r
t
h
e
o
r
t
h
o
g
o
n
a
l
g
r
o
u
p
i
e
O
(
n
)
=
G
L
(
n
,
R
)
[
g
g
=
I
n
a
n
d
T
n
i
s
t
h
e
s
e
t
o
f
s
y
m
m
e
t
r
i
c
d
e
n
i
t
e
p
o
s
i
t
i
v
e
m
a
t
r
i
c
e
s
,
s
u
c
h
t
h
a
t
Q
=
K
Q
.
T
h
i
s
d
e
c
o
m
p
o
s
i
t
i
o
n
o
f
Q
i
s
u
n
i
q
u
e
a
n
d
i
s
r
e
f
e
r
e
d
t
o
a
s
p
o
l
a
r
d
e
c
o
m
p
o
s
i
t
i
o
n
.
W
e
r
e
f
e
r
t
o
F
a
r
a
u
t
(
2
0
0
6
)
f
o
r
f
u
r
t
h
e
r
r
e
s
u
l
t
s
o
n
m
a
t
r
i
x
a
n
a
l
y
s
i
s
.
48