how to forecast on markov regime switch?

[tanghanpost]

Hi,
Very thanks for help.
I'am using month data of 2000m01 to 2014m01, and has built equation using markov regime swithc model, which has dynamic y(t) on c and y(t-1), and has chosed "regime specific error variances", after the estimation, when forecast, eviews 8 always alarm me that "SIGMA is not defined-", I want to know what wrong with the model? and how to forecast on Markov model?
Very thanks.
Tanghan

[EViews Glenn]

Sounds like there might be an issue with EViews. Can you post your workfile? What's the build date on your copy of EViews?

[tanghanpost]

Very thanks for your help.
At first i thought maybe my data has some flaw, so I changed to the data which provided by eviews 8-GNP_hamilton, but the question remains the same. I changed the eqation by dynamic: g c g(-1), the non-switching regressors still are-ar(1) ar(2) ar(3) ar(4), and i chosed the "regime specific error variances", after the estimation, I want to forecast the dependent variable-g, but forecast has teld me that " the SIGMA is not defined". I don't know what wrong with the equation?
Another,
I fined that the forecast dialog is-"SIGMAf" not the variable i want-the dependent variable "gf", then how to use the markov switch model to forecast the dependent variable?
Thank you for spending time to answer me.
Tanghan