EViews.com

Sure! here it goes

It's hard to tell since I'm not that familiar with the model and I can't really see what you are doing in Matlab. I will say that it looks as though the variances aren't settling down, hence the large gradients. Have you compared likelihoods in the two programs at same coefficients? You might also want to try to parameterize your standard deviations in exponential form. I played with that a bit and got better behavior.