Page 1 of 1

Urgent - state space model

Posted: Sun Jul 12, 2009 11:56 pm
by workhard
Hi,

I am trying the State Space Kalman Filter in Eviews to fit a single series.
I encounter a problem that when the series is fitted in a same model but different workfile(the data, equations and setting are exactly the same except I do in different workfile) the results might be very different. What is the reason behind? Is it something to do with the optimisation algolrithms used in Eviews in solving the SS model?

Also, sometimes i get the warning "WARNING: Singular covariance - coefficients are not unique" and the std.error, z-statistics & Prob show "NA". The model below is one of the example that I faced this problem. Can anyone tell me what is the problem?
@signal rklci = c(1)*sv1 + sv2
@state sv1 = c(2)*sv1(-1) + (1-c(2))*sv2(-1) + [var = exp(c(3))]
@state sv2 = c(4)*sv2(-1) + [var = exp(c(5))]

Appreaciate the reply from any helpful person..

Re: Urgent - state space model

Posted: Mon Jul 13, 2009 12:07 am
by trubador
Please search the forum for similar problems.

Re: Urgent - state space model

Posted: Mon Jul 13, 2009 8:18 am
by workhard
Able to get the result after refer to the previous forum..

Thanks a lot :D