EViews.com

Posted: **Wed Apr 28, 2010 2:57 pm**

Hi there,

I'm getting an error message when running the code. Checking out help, I probably misunderstand the SVAR estimation command... The code builds up on VAR model estimated and saved in this workfile in a previous session. Could that be a problem, that one needs to estimate the VAR model and continue with the structural factorization immediately in one and the same session?

Many thanks for helping!!!

Code: Select all

 
 var02.makeresids rev_var02res xpe_var02res gdp_var02res
':: compute cyclically adjusted fiscal policy
            series ca_rev
            ca_rev = rev_var02res - a1(2) * gdp_var02res
            series ca_xpe
            ca_xpe = xpe_var02res - a1(2) * gdp_var02res
':: declare coefficient vectors
            coef(2) c1
            coef(2) c2
            coef(2) b2
':: estimate set of least squares equations
            equation eq_c1_rev.ls gdp_var02res = c1(1) + c1( 2) * ca_rev
            equation eq_c2_xpe.ls gdp_var02res = c2(1) + c2( 2) * ca_xpe
            equation eq_b2_xpe.ls ca_xpe = b2(1) + b2(2) * ca_rev
':: put together the left and right hand side matrix as in B&P (2002)
            matrix(3,3) left
            left.fill(b=r) 1, 0, -a1(2), 0, 1, -b1(2), -c1(2), -c2(2), 1
            matrix(3,3) right
            right.fill(b=r) 1, 0, 0, b2(2), 1, 0, 0, 0, 1
':: estimate structural factorization using the short run matrices left, right
            var02.svar(rtype=patsr,namea=left,nameb=right)

Posted: **Wed Apr 28, 2010 3:39 pm**

Your SVAR doesn't have any parameters to estimate. Both of your matrices (left and right) have been completely filled with values. As explained in the User Guide, you need to use NAs to define which elements are to be estimated.

Posted: **Thu Apr 29, 2010 1:37 am**

Hi QMS Gareth,

thanks for reply. In the code I did not take into account that Structural factorization is strictly an estimation procedure. What I need in this case is rather imposing a specific indentification scheme to come up with structural shocks. The matrices left and right together build the transformation (or projection) matrix which should be used instead of a cholesky factor etc.

My questions:
1. Now I am not sure whether using own transformation matrix for computation of impulse responses is possible in EViews and how to do it?
I doubt that this could be done under Impulse: User specified. By the way from the help documentation it is not completely clear to me, which underlying transformation matrix is used for User specified impulses. Or maybe I just understand it wrongly.

2. Should I try to write a code to compute such impulse responses "by hand"? [I am facing a similar problem with an attempt to impose a combination of short and long run patterns.]

Thanks for helping, it is really urgent for me.

EViews.com

Structural VAR (EV 6)

Structural VAR (EV 6)

Re: Structural VAR (EV 6)

Re: Structural VAR (EV 6)