Page 1 of 1

SSPACE Near_Singular

Posted: Tue Oct 24, 2017 5:49 am
by osindi
Hello,

I'm trying to specify a statespace model and estimate it.
However, it keeps saying that the matrix is near singular.

I'm only using two variables (P in the signal equation which is based on nairu, which is in turn based on the @state equation).

Any guidance would be appreciated

Using Eviews 9.5

Thanks,

Re: SSPACE Near_Singular

Posted: Wed Oct 25, 2017 8:47 am
by EViews Matt
Hello,

It looks like the initial conditions of the estimation happen to produce a singular matrix. Try initializing your coefficients to something other than 0.

Re: SSPACE Near_Singular

Posted: Wed Oct 25, 2017 3:34 pm
by osindi
Hi Matt,

Thanks for the suggestion. Unfortunately even after setting the initial conditions for the state space to nairu0 = 4.5, and nairusd0 = 1 i still get a near singular matrix without much else being said.

It's possible i've not set up the specification correctly, as i'm not sure what else is wrong here.

Any advice would be appreciated,

Thanks

Re: SSPACE Near_Singular

Posted: Wed Oct 25, 2017 4:22 pm
by EViews Matt
Did you try setting the 'c' coefficients so something non-zero as well?

Re: SSPACE Near_Singular

Posted: Thu Oct 26, 2017 12:53 am
by osindi
I only realised after i posted that you meant initialise the parameters, and not the state variable.

That's working now, thanks!