Warning in State Space Estimation

[George]

Dear all,

One more question regarding the estimation of state space models in Eviews. What does the message "WARNING: Singular covariance - coefficients are not unique" imply exactly and what can and must be done in order to eliminate this problem?

Thanks a lot in advance for your feedback!

Regards,
George

[EViews Gareth]

It means that due to the specification of your state space model, there is no unique solution, and EViews is reporting one of the possibilities.

[George]

Dear Gareth,

Thank you so much for yuor reply.

Does this mean that the resolution of this problem can be achieved only through a redefinition of the model only? Or can changes in the initial values (coefficient and/or state) be a key to the solution?

Best,
George

[maragloria]

I'am also encountering the same problem. I have the same question as your last post. What did you do to solve the problem (is it indeed a "problem"?).

Thanks for your time.

[EViews Glenn]

Changes in starting values *can* solve the problem. In some cases, only a modification of the specification will suffice.