EViews.com

im doing a garch(1,1)-m model applied in state space model, my programme is as follows:
@signal r=a+b*r(-1)+d*h+e1
@state h=c(1)+c(2)*h(-1)+c(3)*e1(-1)^2
@state a=a(-1)+e2
@state b=b(-1)+e2
@evar var(e1)=1
@evar var(e2)=exp(c(5))
when i try to run this programme. it says "system contains no nonzero error variance"
can anyone tell me how to fix it?

thank you for your help!

Unfortunately, you cannot conduct a garch-in-mean analysis in EViews' state space object, since at some point you'll have to define squares of error terms or states, which are not allowed due to linearity assumptions.

@trubador

thank you for your help! btw, is there any way that i can estimate my GARCH(1,1)-M state space model? can you please introduce me some other kind of software package?

It would be inappropriate to advise another software package here. However, I think your problem will continue in most cases, since you are trying to put a nonlinear structure in a linear state space model. Either you'll have to find a way to linearize your model, or you'll have to refer to more complex estimation methods (e.g. extended kalman filter, particle filter, etc.).

@trubador

many thanks....