M-GARCH ESTIMATION ERROR

[sam2010]

I'm using the Tse code with up to six variables and wondering if you can help me with this error:

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

*display ' *===============================================================================================';
*display ' * TseCN.prg: Estimates a CC-GARCH model for Canada & perfroms TSE test in RATS 6.3 28/9/09 ';
*display ' * Money and Output with recession dummy split in two ';
*display ' *===============================================================================================';
*
* Tse LM test for constant correlation.
* Tse, Y. K. (2000), “A Test for Constant Correlations in a Multivariate GARCH Model,”
* Journal of Econometrics 98, 107-127.
*
* Updated 08/04 for version 6
*
cal 1960 4 12
all 0 2007:12
open data testweb.xls
data(format=xls,org=obs) / x1 x2 x3 x4 x5 x6

*
*
COMPUTE GSTART=1960:5, GEND=2007:12

*RATS WIZARD
declare symm[series] hhs(6,6)
do i=1,6
do j=1,i
set hhs(i,j) gstart gend = 0.0
end do j
end do i

equation eq1 x1
# constant x2 x3 x4 x5 x6
equation eq2 x2
# constant x2{1} x3{1} x4{1} x5{1} x6{1}
equation eq3 x3
# constant x2{1} x3{1} x4{1} x5{1} x6{1}
equation eq4 x4
# constant x2{1} x3{1} x4{1} x5{1} x6{1}
equation eq5 x5
# constant x2{1} x3{1} x4{1} x5{1} x6{1}
equation eq6 x6
# constant x2{1} x3{1} x4{1} x5{1} x6{1}
*
group garchm eq1 eq2 eq3 eq4 eq5 eq6

garch(model=garchm,p=1,q=1,mv=cc,mvhseries=hhs,pmethod=simplex,piters=50,method=bfgs,robust,iters=100) gstart gend

*Tse prg part
compute n=6
*
dec vect[series] y(n) u(n)
dec vect[frml] resid(n)
set y(1) = x1
set y(2) = x2
set y(3) = x3
set y(4) = x4
set y(5) = x5
set y(6) = x6
*
*
clear u
*
* Do initial regression. Copy initial values for regression parameters
*
LINREG Y(1) GSTART GEND u(1)
# CONSTANT Y(1){1}
COMPUTE B0 = 1.0
COMPUTE B1 = 0.0
COMPUTE B2 = 0.0
COMPUTE B3 = 0.0
COMPUTE B4 = 0.0

* USREC
LINREG Y(2) GSTART GEND u(2)
# CONSTANT Y(2){1} Y(3){1} Y(4){1} Y(5){1} Y(6){1}
COMPUTE B5 = %BETA(1)
COMPUTE B6 = %BETA(2)
COMPUTE B7 = %BETA(3)
COMPUTE B8 = %BETA(4)
COMPUTE B9 = %BETA(5)
COMPUTE B10 = %BETA(6)

LINREG Y(3) GSTART GEND u(3)
# CONSTANT Y(2){1} Y(3){1} Y(4){1} Y(5){1} Y(6){1}
COMPUTE B11 = %BETA(1)
COMPUTE B12 = %BETA(2)
COMPUTE B13 = %BETA(3)
COMPUTE B14 = %BETA(4)
COMPUTE B15 = %BETA(5)
COMPUTE B16 = %BETA(6)


LINREG Y(4) GSTART GEND u(4)
# CONSTANT Y(2){1} Y(3){1} Y(4){1} Y(5){1} Y(6){1}
COMPUTE B17 = %BETA(1)
COMPUTE B18 = %BETA(2)
COMPUTE B19 = %BETA(3)
COMPUTE B20 = %BETA(4)
COMPUTE B21 = %BETA(5)
COMPUTE B22 = %BETA(6)

LINREG Y(5) GSTART GEND u(5)
# CONSTANT Y(2){1} Y(3){1} Y(4){1} Y(5){1} Y(6){1}
COMPUTE B23 = %BETA(1)
COMPUTE B24 = %BETA(2)
COMPUTE B25 = %BETA(3)
COMPUTE B26 = %BETA(4)
COMPUTE B27 = %BETA(5)
COMPUTE B28 = %BETA(6)

LINREG Y(6) GSTART GEND u(6)
# CONSTANT Y(2){1} Y(3){1} Y(4){1} Y(5){1} Y(6){1}
COMPUTE B29 = %BETA(1)
COMPUTE B30 = %BETA(2)
COMPUTE B31 = %BETA(3)
COMPUTE B32 = %BETA(4)
COMPUTE B33 = %BETA(5)
COMPUTE B34 = %BETA(6)
*
NONLIN(parmset=meanparms) B0 B1 B2 B3 B4 B5 B6 B7 B8 B9 B15 B16 B17 B18 $
B19 B20 B21 B22 B23 B24 B25 B26 B27 B28 B29 B30 B31 B32 B33 B34
frml resid(1) = (Y(1)+hhs(1,1)/2-B0*hhs(1,2)-B1*hhs(1,3)-B2*hhs(1,4)-B3*hhs(1,5)-B4*hhs(1,6))
FRML resid(2) = (Y(2)-B5-B6*Y(2){1}-B7*Y(3){1}-B8*Y(4){1}-B9*Y(5){1}-B10*Y(6){1})
FRML resid(3) = (Y(2)-B11-B12*Y(2){1}-B13*Y(3){1}-B14*Y(4){1}-B15*Y(5){1}-B16*Y(6){1})
FRML resid(4) = (Y(2)-B17-B18*Y(2){1}-B19*Y(3){1}-B20*Y(4){1}-B21*Y(5){1}-B22*Y(6){1})
FRML resid(5) = (Y(2)-B23-B24*Y(2){1}-B25*Y(3){1}-B26*Y(4){1}-B27*Y(5){1}-B28*Y(6){1})
FRML resid(6) = (Y(2)-B29-B30*Y(2){1}-B31*Y(3){1}-B32*Y(4){1}-B33*Y(5){1}-B34*Y(6){1})
*print / u
*
* Get the covariance matrix of the residuals.
*
vcv(matrix=rr,print)
# u
*
* h will have the sequence of variance estimates
* uu will have the sequence of uu' matrices
*
dec series[symm] h uu
*
* hx and uux are used when extracting elements from h and uu.
* ux is used when extracting a u vector
*
declare symm hx(n,n) uux(n,n)
declare vect ux(n)
*
* This is used to initialize pre-sample variances.
*
gset h = rr
gset uu = rr
*
*
declare frml[symm] hf
*
frml logl = $
%do(i,1,n,u(i)=resid(i)),$
hx = hf(t) , $
ux = %xt(u,t) , uux = %outerxx(ux),$
h(t) = hx , uu(t) = uux,$
%logdensity(hx,ux)

*
* Constant correlation (reparameterized to use an n x n "symmetric" matrix
* to hold the subdiagonal of the correlation matrix)
*
dec symm qc(n,n)
dec vect vcv(n) vbv(n) vav(n)
nonlin(parmset=garchparms) vcv vbv vav qc


frml hf = %do(i,1,n,%(hx(i,i)=vcv(i)+vav(i)*h{1}(i,i)+vbv(i)*uu{1}(i,i))),$
%do(j,1,n,hx(i,j)=%if(i==j,hx(i,i),qc(i,j)*sqrt(hx(j,j)*hx(i,i)))),hx

ewise qc(i,j)=rr(i,j)/sqrt(rr(i,i)*rr(j,j))
compute vcv=%xdiag(rr),vbv=%const(0.05),vav=%const(0.75)
maximize(parmset=meanparms+garchparms,pmethod=simplex,piters=1,method=bfgs,robusterrors,iters=100,derives=ccderives) logl gstart gend




@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

got : ## SR10. Missing Values And/Or SMPL Options Leave No Usable Data Points

Thanks in advance for your kind help.

[TomDoan]

That's quite an old version of the Tse test; you might want to get the newer one from the web site. At any rate, the nesting on the definition of HF isn't quite right. It needs to be:

Code: Select all

frml hf = %do(i,1,n,%($
     hx(i,i)=vcv(i)+vav(i)*h{1}(i,i)+vbv(i)*uu{1}(i,i),$
     %do(j,1,i,hx(i,j)=%if(i==j,hx(i,i),qc(i,j)*sqrt(hx(j,j)*hx(i,i))))$
     )),hx

The %do(j,....) has to be inside the %do(i,1,n,%(....))

[sam2010]

Thanks a lot Tom,
Yes the code is quite old but I'm mainly interested in the estimation part, since BEKK cannot be used with more than 4 var and I have 6 I tried tse code. Maybe there is a better code !!!

[sam2010]

Hi Tom,
When I try to run the program with the suggested change I got this error:

Code: Select all

Can't Find Match for %(REAL,INTEGER). Closest Match is %(Any)
## SX27. Illegal Combination of Data Types for Operation
>>>>     )<<<<

and a close check show that this error is due to the suggested changes at:

Code: Select all

frml hf = %do(i,1,n,%($
     hx(i,i)=vcv(i)+vav(i)*h{1}(i,i)+vbv(i)*uu{1}(i,i),$
     %do(j,1,i,hx(i,j)=%if(i==j,hx(i,i),qc(i,j)*sqrt(hx(j,j)*hx(i,i))))$
     )),hx

Thanks again for your kind help.

Sam

[TomDoan]

What version of RATS are you using? I don't get that, and I've tried several older versions as well.

[sam2010]

Hi Tom,
I'm using version 6.3. Is it a version compatibility ?

[sam2010]

Thanks for your help.
Sam.

[sam2010]

Hi Tom,
I'M trying to estimate the above model with the wrong covariance matrix. I'm wondering if there is any way of changing the code to account for an off-diag matrix [qc] and diag with ones (cc model):

dec symm qcc(n,n)
dec symm qc(n,n)
dec vect vcv(n) vbv(n) vav(n)
nonlin(parmset=garchparms) vcv vbv vav qc



frml hf = %do(i,1,n,%($
hx(i,i)=vcv(i)+vav(i)*h{1}(i,i)+vbv(i)*uu{1}(i,i),qcc(i,i)=1.0,$
%do(j,1,i,hx(i,j)=%if(i==j,hx(i,i),qcc(i,j)*sqrt(hx(j,j)*hx(i,i) ) ),qcc(i,j)=qc(i,j)/sqrt(qc(i,i)*qc(j,j)))$
)),hx


ewise qc(i,j)=rr(i,j)/sqrt(rr(i,i)*rr(j,j))
ewise qcc(i,j)=qc(i,j)/sqrt(qc(i,i)*qc(j,j))

but the program is not working I believe the qcc part is not working properly.
Thanks for your help.
Sam.

[TomDoan]

I'm not 100% sure of what you're trying to do but

Code: Select all

frml hf = %do(i,1,n,%($
hx(i,i)=vcv(i)+vav(i)*h{1}(i,i)+vbv(i)*uu{1}(i,i),qcc(i,i)=1.0,$
%do(j,1,i,hx(i,j)=%if(i==j,hx(i,i),qcc(i,j)*sqrt(hx(j,j)*hx(i,i) ) ),qcc(i,j)=qc(i,j)/sqrt(qc(i,i)*qc(j,j)))$
)),hx

has what's basically a circular definition of qcc(i,j)---it's using qcc(i,j) in the %do(j,...) segment, then defines qcc(i,j) as the correlation after that.

[sam2010]

Thanks Tom for you reply.
Sam