用S-Plus软件估计出BEKK模型后，如何检验溢出效应啊。从网上找到相应程序并修改如下：

> a.mod=llwbekk$model

> a.mod$arch$value$lag.1[1,2]=0

> a.mod$garch$value$lag.1[1,2]=0

> a.mod$garch$which$lag.1[1,2]=F

> a.mod$arch$which$lag.1[1,2]=F

> a.mod$arch$value$lag.1[2,1]=0

> a.mod$garch$value$lag.1[2,1]=0

> a.mod$arch$which$lag.1[2,1]=F

> a.mod$garch$which$lag.1[2,1]=F

> llwsp=mgarch(series=llw,model=a.mod)

Iteration   0  Step Size = 0.00396202  Likelihood = -1362.52

.....

teration  37  Step Size = 0.196261  Likelihood =  8.35511

> LR.a=-2*(llwsp$likelihood-llwbekk$likelihood)

> LR.a

[1] 4520.734

> 1-pchisq(LR.a,4)

[1] 0

> summary(llwsp)

Warning messages:

  Choleski decomposition not of full rank in: chol(B)

Problem in backsolve(chol(B), diag(k)): x and r should have the same number of rows

Use traceback() to see the call stack

问题：一是LR统计量怎么这么大啊，不论单向还双向的检验没有不显著的，与BEKK模型估计结果不一致；二是如何用Wald统计量进行检验啊?
请高手指点！！！！多谢！！！！！