WARNING: Singular covariance error with Tri-Bekk mgarch

[joburg]

Hi,

Im working on a paper to help model the spillover effects from developed countries to emerging economies. The program uses a Lower-triangular Tri-variate Bekk-mGarch model to help isolate causality. I keep getting a "WARNING: Singular covariance - coefficients are not unique" error.

I looked at the var/cov specifications and they seem to correct. Any help would be greatly appreciated.

Attached is my data sheet as well as the coding.

J

[joburg]

Hi to those experiencing the same problem

Found the problem with the program. It had to do with my OMEGA matrix. For some or other reason, Omega(4) and Omega(5) remained at their initial values = 0. Thus creating a singular matrix in estimation. I've re-specified them as following:

coef(6) omega
omega(1)=(eq1.c(2))^.5
omega(2)=(eq1.c(2))^.5
omega(3)=(eq2.c(2))^.5
omega(4)=(eq2.c(2))^.5
omega(5)=(eq2.c(2))^.5
omega(6)=(eq3.c(2))^.5

But, this is done ad hoc. Any advice on better specifying the initial values to better estimation of the parameters? At the moment i can p-values for my Omega parameters at 0.9998, 0.9999 etc

Hoping for some insight,
J