Dear guys, I write you because I have a problem setting the initial values for state space model. I'm trying to calculate potential GDP from unobserved component following the usual Clark (1987) notation:
@signal log(GDP)*100=sv1+sv2
@state sv1=C(1)+sv1(-1)+ [var = exp(C(2))]
@state sv2=C(3)*sv2(-1)+C(4)*sv3(-1)+ [var = exp(C(5))]
@state sv3=sv2(-1)
The problem is in the initial values for C(1), C(2), C(3), C(4), C(5). There is a theory or a rule which suggest plausible initial values? Anyone can suggest me a procedure to follow? Thanks in advance.
State Space model and initial values
Moderators: EViews Gareth, EViews Moderator
Re: State Space model and initial values
Dear guys, I think to have solved the problem. I have inserted as starting values of C(3) and C(4) (parameters of cycle component) the values of a previous OLS estimation, and for others parameters I have inserted the value 0. All coefficients are statistically significant except for C(5). Only a question: I run the model starting from a date before the available data (that is, I run the model starting from 1947 and my dataset begins in 1960). But this "trick" is fundamental for the convergence. What is the meaning of this result? And it is correct?
I send my dataset with results
I send my dataset with results
- Attachments
-
- sspace_ita.wf1
- (14.31 KiB) Downloaded 494 times
Re: State Space model and initial values
Can anybody help me please?
Who is online
Users browsing this forum: No registered users and 2 guests