我用WINRATS软件编程建模中证指数的基于多元T分布的BEKK模型，观察它们之间的溢出，但是现在编程发现不收敛，尝试了很多的初值都是这样，请问为什么，怎么解决？？？
这个问题很急！希望有大拿解答

*bekk
open data BEKK1.xls
data(format=xls,org=columns) 2 1011 one three
linreg(noprint) one / r1store
#constant
compute a0=%beta(1)
compute w1=%seesq
*
linreg(noprint) three / r2store
#constant
compute b0=%beta(1)
compute w3=%seesq
compute w2=(w1*w3)**0.5
*
vcv(matrix=v,noprint)
# r1store r2store
compute u11store=v(1,1),u12store=v(1,2),$
u22store=v(2,2)
*
set r1 = r1store
set r2 = r2store
set u11 = u11store
set u12 = u12store
set u22 = u22store
nonlin a0 b0 w1 w2 w3 a11 a12 a21 a22 b11 b12 b21 b22
frml h11 = w1 +a11*a11*r1{1}**2+2*a11*a21*(r1{1}*r2{1})+a21*a21*r2{1}**2+b11*b11*u11{1}+2*b11*b21*u12{1}+b21*b21*u22{1}
frml h12 = w2 +a11*a12*r1{1}**2+(a12*a21+a11*a22)*(r1{1}*r2{1})+a21*a21*r2{1}**2+b11*b12*u11{1}+(b12*b21+b11*b22)*u12{1}+b21*b21*u22{1}
frml h22 = w3 +a12*a12*r1{1}**2+2*a12*a22*(r1{1}*r2{1})+a22*a22*r2{1}**2+b12*b12*u11{1}+2*b12*b22*u12{1}+b22*b22*u22{1}
frml resid1 = one - a0
frml resid2 = three - b0
compute a11 = -0.4361
compute a12 = -0.5268
compute a21 = 3.6854e-05
compute a22 = 0.0285
compute b11 = -0.0487
compute b12 = -0.0505
compute b21 = 0.0332
compute b22 = 0.0330
dec symm um
dec vect rv
frml glogl =( u11(t) =h11(t)),( u12(t) = h12(t)),( u22(t) = h22(t)), $
(r1(t) = resid1(t)),(r2(t) = resid2(t)), $
(um =||u11(t)|u12(t),u22(t)||),$
(rv =||r1(t),r2(t)||),$
-0.5*log(%det(um))-0.5*505*(1+(1/1008)*%qform(inv(um),rv))
nlpar(criterion=value,cvcrit=0.00001,subiters=300)
maximize(method=genetic,iters=25,noprint) glogl 4 1011
maximize(method=genetic,robust,recursive,iters=500)  glogl 4 1011