G0041 : Argument must be scalar ERROR

Question

I am not able to execute the program because of this error. When I use another variable then there is no issue. Please help.

The program:

 /*
** THIS PROCEDURE ESTIMATES THE VAR-BEKK-GARCH-M MODEL;
**
*/
new;
cls;
@===========call different library=============@
library maxlik, pgraph;
#include maxlik.ext;
maxset;
graphset;
@===============================================@
format /ld 8,6;
@==================set output file==============@
output file=varmgarchm_manuf_lag2.out reset;
@===============================================@
 
@============= load data file===================@
nor0=358; @=====number of rows=================@
noc=2; @========number of columns===========@
//load data[nor0,noc]=ukbr.txt;
 
data1mg = loadd("A:/1_research21/1.1_obj1/1_MGARCH_part/analysis/model1/manuf/data1mg.xlsx", "rt + manufgrowth");
 
z0=data1mg[.,"rt" "manufgrowth"];@==level values========@
//z0=(ln(z0[2:nor0,.])-ln(z0[1:nor0-1,.]))*100;
nor=rows(z0);
@==========================================@
n=noc; @===number of equations/number of dependent variable=======@
p=1; @====lags in VAR part=======@
q=1; @====lags in ARCH part======@
@===============================================@
/* Procedure definition area */
 
/********************** PROC VARLAGS *****************************
** last update: 5 Dec 95 previous: 15 June 94
** AUTHOR
** Alan G. Isaac
** FORMAT
** {x,xlags} = varlags(var,lags)
** INPUT
** var - T x K matrix
** lags - scalar, number of lags of var (a positive integer)
** OUTPUT
** x - (T - lags) x K matrix, the last T-lags rows of var
** xlags - (T - lags) x lags*cols(var) matrix,
** being the 1st through lags-th
** values of var corresponding to the values in x
** i.e, the appropriate rows of x(-1)~x(-2)~etc.
** GLOBAL VARIABLES: none
**********************************************************************/
proc(2)=varlags(var,lags);
local xlags;
xlags = shiftr((ones(1,lags) .*. var)',seqa(1-lags,1,lags)
.*. ones(cols(var),1),miss(0,0))';
retp(trimr(var,lags,0),trimr(xlags,0,lags));
endp;
@==========================================@
{y,ylag}=varlags(z0,p);
t=rows(y~ylag);
 
c=ones(t,1); @===========constant in VAR ============== @
@=========================================================@
data=y~c~ylag;
/* **************set initial parameters***************** */
/*theta=ones(23,1).*0;
theta[1:2]=0.1|0.1; theta[3:6]=ones(4,1).*0.06; theta[7:12]=ones(6,1).*0.13;
theta[13:15]=ones(3,1).*0.03; theta[16:19]=ones(4,1).*0.14; theta[20:23]=ones(4,1).*0.7; */
theta={0.0056
-0.0169
-0.0702
0.2558
0.0114
0.0921
0.0414
-0.0382
-0.0153
0.1139
0.0023
0.0446
-0.0954
0.54
0.9092
0.2554
0.0158
0.1033
0.6264
0.9065
-0.008
0.1609
0.3898
};
 
/* ******************************************************* */
@======================cml setting=======================@
__title="VAR WITH BEKK-GARCH IN MEAN MODEL UK-BR_LAG1";
_max_DirTol=1.e-5;
_max_Algorithm=2;
_max_CovPar=0;
_max_MaxIters=1000;
__output=1;
/*_cml_Bounds=*/
/*_cml_ParNames="b10"|(0$+"b1"$+ftocv(seqa(1,1,cols(lagp)*neqn),1,0))|
"b20"|(0$+"b2"$+ftocv(seqa(1,1,cols(lagp)*neqn),1,0))|
"b30"|(0$+"b3"$+ftocv(seqa(1,1,cols(lagp)*neqn),1,0));*/
/* ********************************************************************** */
{th,f,g,covmat,retcode}=maxprt(maxlik(data,0,&varmgarchmll,theta));
/* ********************************************************************** */
/* ***************************fit of the model********************************* */
{yhatt,et,est,Ht}=varmgarchmfit(th,data);
 
"";
//"fit1 fit2 err1 err2 stderr1 stderr2 h1 h2";
//yhatt~et~est~Ht;
@================== log likeltihood value================@
llv=f*t;
"Loglikelihood Value";
llv;
"number of observation";
t;
/* ************* likelihood ratio test ************* */
llv1=-9110.751367;
np1=17;
 
lr=-2*(llv1-llv);
lrpval=cdfchic(lr,(rows(th)-np1));
"LIKELIHOOD RATIO TEEST";"";
"LR LRPVAL";
lr~lrpval;
/* ******************** AIC/BIC ***************** */
@====================== AIC/BIC===================@
"==========================================";
"AIC BIC";"";
 
AIC= -2*f+2*rows(th)/t;
BIC=-2*f+2*rows(th)*ln(ln(t))/t;
 
AIC~BIC;
 
@===============residual test=========@
varres=est;
smax=10;
Numvars = 2;
lags=1; /* v is the number of variables in the VAR model. */
INVCOV00 = INVPD(((varres[1:T,.])'*(varres[1:T,.]))/(T));
LBS = 0;
teststat = zeros(1,Smax);
pvalu = zeros(1,Smax);
S=1;
do until S > Smax;
{LBS} = LBPort (S, varres,T,LBS,INVCOV00);
LBSpv = cdfchic(LBS,numvars^2*(S-lags));
teststat[.,S] = LBS;
pvalu[.,S] = LBSpv;
S = S +1;
endo;
"teststat";
teststat;
"pvalu";
pvalu;
/* ******************** IN-SAMPLE RESULT *************** */
RMSE=sqrt(meanc(et.^2));
"RMSE";
RMSE;
 
 
 
/* *******************plot *************** */
/* Make graphs of actual and fitted series, transition function over time,
residuals and transition function versus transition variable */
graphset;
_pdate="";
_pnum=2;
_ptitlht=0.3;
_paxht=0.15;
_ptek="varmgarchm_ukbr_lag1.tkf";
begwind;
window(3,2,0);
_ptitle="Series1+Fit1";
xy(0,data[.,1]~yhatt[.,1]);
nextwind;
_ptitle="Series2+Fit2";
xy(0,data[.,2]~yhatt[.,2]);
nextwind;
_ptitle="H1 vs. time";
xy(0,ht[.,1]);
nextwind;
_ptitle="H2 vs. time";
plotxy(seqa(1,1,rows(ht)),ht[.,2]);
nextwind;
_plctrl = 0;
_ptitle="H1 vs e1(t-1)";
temp1=sortc(et[1:t-1,1]~ht[2:t,1],1);
xy(temp1[.,1],temp1[.,2]);
nextwind;
_plctrl = 0 ;
_ptitle="H2 vs e2(t-1)";
temp2=sortc(et[1:t-1,2]~ht[2:t,2],1);
xy(temp2[.,1],temp2[.,2]);
endwind;
 
 
 
/********************** PROCEDURE DEFINITION AREA ************************** */
 
/* ************* computes the likelihood function ************************ */
proc(1)=varmgarchmll(theta,data);
local tc,tn,y1,y,x1,x,tb,b1,b,d1,d,o1,o2,o,a1,a,bg1,bg,i,m,ht,e,za,oldt,ll;
tc=cols(data);
tn=rows(data);
y1=data[.,1:n];
x1=data[.,n+1:tc];
@===========CREATING PARAMETER MATRIX========================@
tb=rows(theta);
b1=theta[1:n*(1+p*n)]; @=====var parameters in vech=======@
b=reshape(b1,(1+p*n),n)'; @=====var parameters=======@
 
d1=theta[n*(1+p*n)+1:(n*(1+p*n)+n*(n*(n+1)/2))]; @====in-mean parameter in vech=====@
d=reshape(d1,(n*(n+1)/2),n)'; @============in-mean parameter================@
 
o1=theta[(n*(1+p*n)+n*(n*(n+1)/2))+1:(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)];
o2=xpnd(o1);
o=lowmat(o2);
 
a1=theta[(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)+1:(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)+n^2];
a=reshape(a1,n,n);
 
bg1=theta[(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)+n^2+1:tb];
bg=reshape(bg1,n,n);
 
@============= COMPLETE PARAMETER MATRIX==========================@
 
y=y1';
x=x1';
/* ************ initialize the matrics ************************ */
 
m=zeros(n,tn);
ht=arrayinit(tn|n|n,0);
e=zeros(n,tn);
ll=zeros(tn,1);
@=====================================@
/* ************* 1st observation of the matrics ****************** */
m[.,1]=meanc(y');
e[.,1]=y[.,1]-m[.,1];
ht[1,.,.]=vcx(y1);
/******************************************************** */
i=2;
do while i<=tn;
ht[i,.,.]=o*o'+a*(e[.,i-1]*e[.,i-1]')*a'+bg*(arraytomat(ht[i-1,.,.]))*bg';
m[.,i]=b*x[.,i]+d*(vech(arraytomat(ht[i,.,.])));
e[.,i]=y[.,i]-m[.,i];
oldt=trapchk(1);
trap 1,1;
za=solpd(e[.,i],arraytomat(ht[i,.,.]));
trap oldt,1;
if not scalmiss(za);
 
ll[i]=-0.5*n*ln(2*pi)-0.5*ln(det(arraytomat(ht[i,.,.])))-0.5*(e[.,i]')*za;
else ;
ll[i]=error(0);
 
endif;
i=i+1;
endo;
retp(ll);
endp;
 
@==========================================================@
 
@================ fitting the model==========================@
/* ************* COMPUTES THE IN-SAMPLE FIITED VALUE ************************ */
proc(4)=varmgarchmfit(theta,data);
local tc,tn,y1,y,x1,x,tb,b1,b,d1,d,o1,o2,o,a1,a,bg1,bg,i,m,ht,e,hht,est;
tc=cols(data);
tn=rows(data);
y1=data[.,1:n];
x1=data[.,n+1:tc];
@===========CREATING PARAMETER MATRIX========================@
tb=rows(theta);
b1=theta[1:n*(1+p*n)]; @=====var parameters in vech=======@
b=reshape(b1,(1+p*n),n)'; @=====var parameters=======@
 
d1=theta[n*(1+p*n)+1:(n*(1+p*n)+n*(n*(n+1)/2))]; @====in-mean parameter in vech=====@
d=reshape(d1,(n*(n+1)/2),n)'; @============in-mean parameter================@
 
o1=theta[(n*(1+p*n)+n*(n*(n+1)/2))+1:(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)];
o2=xpnd(o1);
o=lowmat(o2);
 
a1=theta[(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)+1:(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)+n^2];
a=reshape(a1,n,n);
 
bg1=theta[(n*(1+p*n)+n*(n*(n+1)/2))+(n*(n+1)/2)+n^2+1:tb];
bg=reshape(bg1,n,n);
 
@============= COMPLETE PARAMETER MATRIX==========================@
 
y=y1';
x=x1';
/* ************ initialize the matrics ************************ */
 
m=zeros(n,tn);
ht=arrayinit(tn|n|n,0);
hht=zeros(n,tn);
e=zeros(n,tn);
est=zeros(n,tn);
@=====================================@
/* ************* 1st observation of the matrics ****************** */
 
ht[1,.,.]=vcx(y1);
hht[.,1]=diag(arraytomat(ht[1,.,.]));
m[.,1]=meanc(y');
e[.,1]=y[.,1]-m[.,1];
est[.,1]=e[.,1]./hht[.,1];
 
/******************************************************** */
i=2;
do while i<=tn;
ht[i,.,.]=o*o'+a*(e[.,i-1]*e[.,i-1]')*a'+bg*(arraytomat(ht[i-1,.,.]))*bg';
hht[.,i]=diag(arraytomat(ht[i,.,.]));
m[.,i]=b*x[.,i]+d*(vech(arraytomat(ht[i,.,.])));
e[.,i]=y[.,i]-m[.,i];
est[.,i]=e[.,1]./hht[.,i];
i=i+1;
endo;
retp(m',e',est',hht');
endp;
 
/*************************************************************************
----proc LBPort
----AUTHOR: Scott Hacker (in cooperation with Hatemi-J)
----INPUT:
S - order of autocorrelation to test for
varres - residuals from VAR regression
T - number of observations
LBS - LBS statistic for S -1
INVCOV00 - var-cov matrix for residuals
----OUTPUT: Ljeung-Box Statistic:
LBS = T(T) x sum (j=1 to S) of (1/(T-j)) x trace(C(0,j)inv[C(0,0)]C'(0,j)inv[C(0,0)])
where C(0,j) = sum(t = j+1 to T) for e(t)e'(t-j), and
e(t) is the residual from the estimation of
y(t) = v + A(1)y(t-1) + . . . + A(k)y(t-k) + error(t)
Note this definition is used rather than
LBS = T(T+2) x sum (j=1 to S) of (1/(T-j)) x trace(C(0,j)inv[C(0,0)]C'(0,j)inv[C(0,0)])
as the first definition matches equation 4.4.23 in Lutkepohl's textbook on
Multivariate Time Series.
----GLOBAL VARIABLES: none
----NB: none.
************************************************************************/
proc (1) = LBPort(S, varres,T,LBS,INVCOV00);
local cov0S,tracematrix,pvalue, RESNOW, RESLAGGED, j, COV0J;
COV0S = ( (varres[S+1:T,.])'*(varres[1:T-S,.]) )/T;
j = S;
RESNOW = varRES[j+1:T,.];
/*print "line 662 RESNOW=";; RESNOW;*/
RESLAGGED = varRES[1:T-j,.];
COV0j = (RESNOW'*RESLAGGED)/T;
/*format /rd 12,7; print "line 654 COV0S";; COV0S;" S=";; S; "COV0J=";; cov0j;*/
tracematrix = COV0S*INVCOV00*COV0S'*INVCOV00;
LBS = LBS + (T*(T)/(T-S))*ones(1,rows(tracematrix))*diag(tracematrix);
/* print " line 667! LBS=";LBS; */
retp(LBS);
endp;

ERROR::

G0041 : Argument must be scalar [maxlik.src, line 767]
Currently active call:
File maxlik.src, line 767, in maxprt
print "Mean log-likelihood " ftos(f,"%#.lG",15,6);
Traceback:
File mgarch1.gss, line 111, in <main>
{th,f,g,covmat,retcode}=maxprt(maxlik(data,0,&varmgarchmll,theta));

1 Answer

Your Answer

admin · Answer 1

The most likely problem is that the return value from varmgarchmll() is a complex number.

You can check this using the iscplx() procedure. Then if it is complex, you can check to see if the complex data is just round-off error with the hasimag() procedure. You can do that by changing the last part of varmgarchmll() from:

         i=i+1;
    endo;
    retp(ll);
endp;

to

         i=i+1;
    endo;
 
    // Is the log-likelihood value complex?
    if iscplx(ll);
        // Is the complex part, just round-off error?
        if hasimag(ll);
            print "The log likelihood is complex:";
            print "ll = " ll;
            print "The current parameters are:";
            print theta;
        else;
            // If the complex part is just round-off, remove it
            ll = real(ll);
        endif;
    endif;
    retp(ll);
endp;

link

admin

47

admin · Answer 2

The most likely problem is that the return value from varmgarchmll() is a complex number.

You can check this using the iscplx() procedure. Then if it is complex, you can check to see if the complex data is just round-off error with the hasimag() procedure. You can do that by changing the last part of varmgarchmll() from:

         i=i+1;
    endo;
    retp(ll);
endp;

to

         i=i+1;
    endo;
 
    // Is the log-likelihood value complex?
    if iscplx(ll);
        // Is the complex part, just round-off error?
        if hasimag(ll);
            print "The log likelihood is complex:";
            print "ll = " ll;
            print "The current parameters are:";
            print theta;
        else;
            // If the complex part is just round-off, remove it
            ll = real(ll);
        endif;
    endif;
    retp(ll);
endp;

link

admin

47

G0041 : Argument must be scalar ERROR

1 Answer

Your Answer

1 Answer

You must login to post answers.

Have a Specific Question?

Need Support?

	/*
	** THIS PROCEDURE ESTIMATES THE VAR-BEKK-GARCH-M MODEL;
	**
	*/
	new;
	cls;
	@===========call different library=============@
	library maxlik, pgraph;
	#include maxlik.ext;
	maxset;
	graphset;
	@===============================================@
	format /ld 8,6;
	@==================set output file==============@
	output file=varmgarchm_manuf_lag2.out reset;
	@===============================================@

	@============= load data file===================@
	nor0=358; @=====number of rows=================@
	noc=2; @========number of columns===========@
	//load data[nor0,noc]=ukbr.txt;

	data1mg = loadd("A:/1_research21/1.1_obj1/1_MGARCH_part/analysis/model1/manuf/data1mg.xlsx", "rt + manufgrowth");

	z0=data1mg[.,"rt" "manufgrowth"];@==level values========@
	//z0=(ln(z0[2:nor0,.])-ln(z0[1:nor0-1,.]))*100;
	nor=rows(z0);
	@==========================================@
	n=noc; @===number of equations/number of dependent variable=======@
	p=1; @====lags in VAR part=======@
	q=1; @====lags in ARCH part======@
	@===============================================@
	/* Procedure definition area */

	/******************** PROC VARLAGS ***************************
	** last update: 5 Dec 95 previous: 15 June 94
	** AUTHOR
	** Alan G. Isaac
	** FORMAT
	** {x,xlags} = varlags(var,lags)
	** INPUT
	** var - T x K matrix
	** lags - scalar, number of lags of var (a positive integer)
	** OUTPUT
	** x - (T - lags) x K matrix, the last T-lags rows of var
	** xlags - (T - lags) x lags*cols(var) matrix,
	** being the 1st through lags-th
	** values of var corresponding to the values in x
	** i.e, the appropriate rows of x(-1)~x(-2)~etc.
	** GLOBAL VARIABLES: none
	**********************************************************************/
	proc(2)=varlags(var,lags);
	local xlags;
	xlags = shiftr((ones(1,lags) .*. var)',seqa(1-lags,1,lags)
	.*. ones(cols(var),1),miss(0,0))';
	retp(trimr(var,lags,0),trimr(xlags,0,lags));
	endp;
	@==========================================@
	{y,ylag}=varlags(z0,p);
	t=rows(y~ylag);

	c=ones(t,1); @===========constant in VAR ============== @
	@=========================================================@
	data=y~c~ylag;
	/* ************set initial parameters*************** */
	/theta=ones(23,1).0;
	theta[1:2]=0.1\|0.1; theta[3:6]=ones(4,1).0.06; theta[7:12]=ones(6,1).0.13;
	theta[13:15]=ones(3,1).0.03; theta[16:19]=ones(4,1).0.14; theta[20:23]=ones(4,1).0.7; /
	theta={0.0056
	-0.0169
	-0.0702
	0.2558
	0.0114
	0.0921
	0.0414
	-0.0382
	-0.0153
	0.1139
	0.0023
	0.0446
	-0.0954
	0.54
	0.9092
	0.2554
	0.0158
	0.1033
	0.6264
	0.9065
	-0.008
	0.1609
	0.3898
	};

	/* ******************************************************* */
	@======================cml setting=======================@
	__title="VAR WITH BEKK-GARCH IN MEAN MODEL UK-BR_LAG1";
	_max_DirTol=1.e-5;
	_max_Algorithm=2;
	_max_CovPar=0;
	_max_MaxIters=1000;
	__output=1;
	/_cml_Bounds=/
	/_cml_ParNames="b10"\|(0$+"b1"$+ftocv(seqa(1,1,cols(lagp)neqn),1,0))\|
	"b20"\|(0$+"b2"$+ftocv(seqa(1,1,cols(lagp)*neqn),1,0))\|
	"b30"\|(0$+"b3"$+ftocv(seqa(1,1,cols(lagp)neqn),1,0));/
	/* ********************************************************************** */
	{th,f,g,covmat,retcode}=maxprt(maxlik(data,0,&varmgarchmll,theta));
	/* ********************************************************************** */
	/* *************************fit of the model******************************* */
	{yhatt,et,est,Ht}=varmgarchmfit(th,data);

	"";
	//"fit1 fit2 err1 err2 stderr1 stderr2 h1 h2";
	//yhatt~et~est~Ht;
	@================== log likeltihood value================@
	llv=f*t;
	"Loglikelihood Value";
	llv;
	"number of observation";
	t;
	/* *********** likelihood ratio test *********** */
	llv1=-9110.751367;
	np1=17;

	lr=-2*(llv1-llv);
	lrpval=cdfchic(lr,(rows(th)-np1));
	"LIKELIHOOD RATIO TEEST";"";
	"LR LRPVAL";
	lr~lrpval;
	/* ****************** AIC/BIC *************** */
	@====================== AIC/BIC===================@
	"==========================================";
	"AIC BIC";"";

	AIC= -2f+2rows(th)/t;
	BIC=-2f+2rows(th)*ln(ln(t))/t;

	AIC~BIC;

	@===============residual test=========@
	varres=est;
	smax=10;
	Numvars = 2;
	lags=1; /* v is the number of variables in the VAR model. */
	INVCOV00 = INVPD(((varres[1:T,.])'*(varres[1:T,.]))/(T));
	LBS = 0;
	teststat = zeros(1,Smax);
	pvalu = zeros(1,Smax);
	S=1;
	do until S > Smax;
	{LBS} = LBPort (S, varres,T,LBS,INVCOV00);
	LBSpv = cdfchic(LBS,numvars^2*(S-lags));
	teststat[.,S] = LBS;
	pvalu[.,S] = LBSpv;
	S = S +1;
	endo;
	"teststat";
	teststat;
	"pvalu";
	pvalu;
	/* ****************** IN-SAMPLE RESULT ************* */
	RMSE=sqrt(meanc(et.^2));
	"RMSE";
	RMSE;



	/* *****************plot ************* */
	/* Make graphs of actual and fitted series, transition function over time,
	residuals and transition function versus transition variable */
	graphset;
	_pdate="";
	_pnum=2;
	_ptitlht=0.3;
	_paxht=0.15;
	_ptek="varmgarchm_ukbr_lag1.tkf";
	begwind;
	window(3,2,0);
	_ptitle="Series1+Fit1";
	xy(0,data[.,1]~yhatt[.,1]);
	nextwind;
	_ptitle="Series2+Fit2";
	xy(0,data[.,2]~yhatt[.,2]);
	nextwind;
	_ptitle="H1 vs. time";
	xy(0,ht[.,1]);
	nextwind;
	_ptitle="H2 vs. time";
	plotxy(seqa(1,1,rows(ht)),ht[.,2]);
	nextwind;
	_plctrl = 0;
	_ptitle="H1 vs e1(t-1)";
	temp1=sortc(et[1:t-1,1]~ht[2:t,1],1);
	xy(temp1[.,1],temp1[.,2]);
	nextwind;
	_plctrl = 0 ;
	_ptitle="H2 vs e2(t-1)";
	temp2=sortc(et[1:t-1,2]~ht[2:t,2],1);
	xy(temp2[.,1],temp2[.,2]);
	endwind;



	/******************** PROCEDURE DEFINITION AREA ************************ */

	/* *********** computes the likelihood function ********************** */
	proc(1)=varmgarchmll(theta,data);
	local tc,tn,y1,y,x1,x,tb,b1,b,d1,d,o1,o2,o,a1,a,bg1,bg,i,m,ht,e,za,oldt,ll;
	tc=cols(data);
	tn=rows(data);
	y1=data[.,1:n];
	x1=data[.,n+1:tc];
	@===========CREATING PARAMETER MATRIX========================@
	tb=rows(theta);
	b1=theta[1:n(1+pn)]; @=====var parameters in vech=======@
	b=reshape(b1,(1+p*n),n)'; @=====var parameters=======@

	d1=theta[n(1+pn)+1:(n(1+pn)+n(n(n+1)/2))]; @====in-mean parameter in vech=====@
	d=reshape(d1,(n*(n+1)/2),n)'; @============in-mean parameter================@

	o1=theta[(n(1+pn)+n(n(n+1)/2))+1:(n(1+pn)+n(n(n+1)/2))+(n*(n+1)/2)];
	o2=xpnd(o1);
	o=lowmat(o2);

	a1=theta[(n(1+pn)+n(n(n+1)/2))+(n(n+1)/2)+1:(n(1+pn)+n(n(n+1)/2))+(n(n+1)/2)+n^2];
	a=reshape(a1,n,n);

	bg1=theta[(n(1+pn)+n(n(n+1)/2))+(n*(n+1)/2)+n^2+1:tb];
	bg=reshape(bg1,n,n);

	@============= COMPLETE PARAMETER MATRIX==========================@

	y=y1';
	x=x1';
	/* ********** initialize the matrics ********************** */

	m=zeros(n,tn);
	ht=arrayinit(tn\|n\|n,0);
	e=zeros(n,tn);
	ll=zeros(tn,1);
	@=====================================@
	/* *********** 1st observation of the matrics **************** */
	m[.,1]=meanc(y');
	e[.,1]=y[.,1]-m[.,1];
	ht[1,.,.]=vcx(y1);
	/******************************************************** */
	i=2;
	do while i<=tn;
	ht[i,.,.]=oo'+a(e[.,i-1]e[.,i-1]')a'+bg(arraytomat(ht[i-1,.,.]))bg';
	m[.,i]=bx[.,i]+d(vech(arraytomat(ht[i,.,.])));
	e[.,i]=y[.,i]-m[.,i];
	oldt=trapchk(1);
	trap 1,1;
	za=solpd(e[.,i],arraytomat(ht[i,.,.]));
	trap oldt,1;
	if not scalmiss(za);

	ll[i]=-0.5nln(2pi)-0.5ln(det(arraytomat(ht[i,.,.])))-0.5(e[.,i]')za;
	else ;
	ll[i]=error(0);

	endif;
	i=i+1;
	endo;
	retp(ll);
	endp;

	@==========================================================@

	@================ fitting the model==========================@
	/* *********** COMPUTES THE IN-SAMPLE FIITED VALUE ********************** */
	proc(4)=varmgarchmfit(theta,data);
	local tc,tn,y1,y,x1,x,tb,b1,b,d1,d,o1,o2,o,a1,a,bg1,bg,i,m,ht,e,hht,est;
	tc=cols(data);
	tn=rows(data);
	y1=data[.,1:n];
	x1=data[.,n+1:tc];
	@===========CREATING PARAMETER MATRIX========================@
	tb=rows(theta);
	b1=theta[1:n(1+pn)]; @=====var parameters in vech=======@
	b=reshape(b1,(1+p*n),n)'; @=====var parameters=======@

	d1=theta[n(1+pn)+1:(n(1+pn)+n(n(n+1)/2))]; @====in-mean parameter in vech=====@
	d=reshape(d1,(n*(n+1)/2),n)'; @============in-mean parameter================@

	o1=theta[(n(1+pn)+n(n(n+1)/2))+1:(n(1+pn)+n(n(n+1)/2))+(n*(n+1)/2)];
	o2=xpnd(o1);
	o=lowmat(o2);

	a1=theta[(n(1+pn)+n(n(n+1)/2))+(n(n+1)/2)+1:(n(1+pn)+n(n(n+1)/2))+(n(n+1)/2)+n^2];
	a=reshape(a1,n,n);

	bg1=theta[(n(1+pn)+n(n(n+1)/2))+(n*(n+1)/2)+n^2+1:tb];
	bg=reshape(bg1,n,n);

	@============= COMPLETE PARAMETER MATRIX==========================@

	y=y1';
	x=x1';
	/* ********** initialize the matrics ********************** */

	m=zeros(n,tn);
	ht=arrayinit(tn\|n\|n,0);
	hht=zeros(n,tn);
	e=zeros(n,tn);
	est=zeros(n,tn);
	@=====================================@
	/* *********** 1st observation of the matrics **************** */

	ht[1,.,.]=vcx(y1);
	hht[.,1]=diag(arraytomat(ht[1,.,.]));
	m[.,1]=meanc(y');
	e[.,1]=y[.,1]-m[.,1];
	est[.,1]=e[.,1]./hht[.,1];

	/******************************************************** */
	i=2;
	do while i<=tn;
	ht[i,.,.]=oo'+a(e[.,i-1]e[.,i-1]')a'+bg(arraytomat(ht[i-1,.,.]))bg';
	hht[.,i]=diag(arraytomat(ht[i,.,.]));
	m[.,i]=bx[.,i]+d(vech(arraytomat(ht[i,.,.])));
	e[.,i]=y[.,i]-m[.,i];
	est[.,i]=e[.,1]./hht[.,i];
	i=i+1;
	endo;
	retp(m',e',est',hht');
	endp;

	/*************************************************************************
	----proc LBPort
	----AUTHOR: Scott Hacker (in cooperation with Hatemi-J)
	----INPUT:
	S - order of autocorrelation to test for
	varres - residuals from VAR regression
	T - number of observations
	LBS - LBS statistic for S -1
	INVCOV00 - var-cov matrix for residuals
	----OUTPUT: Ljeung-Box Statistic:
	LBS = T(T) x sum (j=1 to S) of (1/(T-j)) x trace(C(0,j)inv[C(0,0)]C'(0,j)inv[C(0,0)])
	where C(0,j) = sum(t = j+1 to T) for e(t)e'(t-j), and
	e(t) is the residual from the estimation of
	y(t) = v + A(1)y(t-1) + . . . + A(k)y(t-k) + error(t)
	Note this definition is used rather than
	LBS = T(T+2) x sum (j=1 to S) of (1/(T-j)) x trace(C(0,j)inv[C(0,0)]C'(0,j)inv[C(0,0)])
	as the first definition matches equation 4.4.23 in Lutkepohl's textbook on
	Multivariate Time Series.
	----GLOBAL VARIABLES: none
	----NB: none.
	************************************************************************/
	proc (1) = LBPort(S, varres,T,LBS,INVCOV00);
	local cov0S,tracematrix,pvalue, RESNOW, RESLAGGED, j, COV0J;
	COV0S = ( (varres[S+1:T,.])'*(varres[1:T-S,.]) )/T;
	j = S;
	RESNOW = varRES[j+1:T,.];
	/print "line 662 RESNOW=";; RESNOW;/
	RESLAGGED = varRES[1:T-j,.];
	COV0j = (RESNOW'*RESLAGGED)/T;
	/format /rd 12,7; print "line 654 COV0S";; COV0S;" S=";; S; "COV0J=";; cov0j;/
	tracematrix = COV0SINVCOV00COV0S'*INVCOV00;
	LBS = LBS + (T(T)/(T-S))ones(1,rows(tracematrix))*diag(tracematrix);
	/* print " line 667! LBS=";LBS; */
	retp(LBS);
	endp;

	i=i+1;
	endo;

	// Is the log-likelihood value complex?
	if iscplx(ll);
	// Is the complex part, just round-off error?
	if hasimag(ll);
	print "The log likelihood is complex:";
	print "ll = " ll;
	print "The current parameters are:";
	print theta;
	else;
	// If the complex part is just round-off, remove it
	ll = real(ll);
	endif;
	endif;
	retp(ll);
	endp;