Page 1 of 1

Parsing state space gets stuck

Posted: Mon Oct 21, 2013 3:54 am
by JamesK
Hi,

I enter a specification which models the behavior of futures with growth at different horizons as state variables (similar to Jegadeesh & Pennacchi 1996). Running the model results in Eviews displaying "parsing specification", but essentially it gets stuck and non-responsive right away. Only by switching Eviews off can this be resolved. I varied the convergence criteria and the number of observations - to no avail.

Could anyone please help out?

The model specification is as follows:

param c(1) 0.01 c(2) 0.01 c(3) 0.01 c(4) 0.001

@signal f11 = c(1)+(z1-(c(3)+c(2)*z2)/(c(2)-c(3)))*exp(-c(2))+(1+z2)*c(2)/(c(2)-c(3))*exp(-c(3))+[var = exp(c(4))]
@signal f12 = 2*c(1)+(z1-(c(3)+c(2)*z2)/(c(2)-c(3)))*(exp(-c(2))+exp(-2*c(2)))+(1+z2)*c(2)/(c(2)-c(3))*(exp(-c(3))+exp(-2*c(3)))+[var = exp(c(4))]
@signal f13 = 3*c(1)+(z1-(c(3)+c(2)*z2)/(c(2)-c(3)))*(exp(-c(2))+exp(-2*c(2))+exp(-3*c(2)))+(1+z2)*c(2)/(c(2)-c(3))*(exp(-c(3))+exp(-2*c(3))+exp(-3*c(3)))+[var = exp(c(4))]
@signal f14 = 4*c(1)+(z1-(c(3)+c(2)*z2)/(c(2)-c(3)))*(exp(-c(2))+exp(-2*c(2))+exp(-3*c(2))+exp(-4*c(2)))+(1+z2)*c(2)/(c(2)-c(3))*(exp(-c(3))+exp(-2*c(3))+exp(-3*c(3))+exp(-4*c(3)))+[var = exp(c(4))]

@state z1=(1+(c(3)*exp(-c(2))-c(2)*exp(-c(3)))/(c(2)-c(3)))*c(1)+exp(-c(2))*z1(-1)+c(2)/(c(2)-c(3))*(exp(-c(3))-exp(-c(2)))*z2(-1)+[var = exp(c(4))]
@state z2=(1-exp(-c(3)))*c(1)+0*z1(-1)+exp(-c(3))*z2(-1)+[var = exp(c(4))]

Re: Parsing state space gets stuck

Posted: Mon Oct 21, 2013 5:40 am
by trubador
It is difficult to detect the source of the problem without seeing the data (or workfile). Try assigning different (also from each other) starting values for coefficients, since you perform arithmetic operations that yield to division by zero at the very beginning of iteration. And make sure that whether all signal and state variables have the same variance.

Re: Parsing state space gets stuck

Posted: Mon Oct 21, 2013 6:12 am
by JamesK
Thank you.

I have uploaded the file with the data (only f11-f14 are used)

Different starting values always results in the same problem: Eviews gets jammed straight away. And all variables do indeed have the same variance.

Since the system gets jammed and there is no error message, does this signify anything?

Re: Parsing state space gets stuck

Posted: Mon Oct 21, 2013 6:50 am
by trubador
The following starting values work fine:

Code: Select all

param c(1) 0.03 c(2) 0.02 c(3) 0.01 c(4) 0.001
Please also check the build date of your copy of EViews and update if necessary.