EViews.com

I am using the following to describe short-run restrictions to a VAR:

@e1 = C(1)*@e1 + C(2)*@u1
@e2 = C(3)*@e1 + C(4)*@u2
@e3 = C(5)*@e2 + C(6)*@u3

to describe a decomposition matrix:
110
110
011

applied to a vector :
A
B
C

in other words:
• A responds to shocks to itself, and B
• B responds to shocks to itself, and A
• C responds to shocks to itself, and A

I'm getting an error "Hessian of structural VAR is singular at starting values." I'm new to time-series work, and haven't found anything in the user guide about this particular error. Is the problem something to do with my code, or something else?

Your help is very much appreciated.

Thanks.
Christian

Christian

Hi ckaas,

skimming your post, in your 1st equation

@e1 = C(1)*@e1 + C(2)*@u1

you force @e1 to depend on itsef what you 'd never do in an equation estimation.
You confuse yourself by writing the transformation scheme as

110
110
011

which is completely wrong. To my impression you strongly need to read some econometric textbook to understand the disctinction and relationship between reduced form residuals and structural shocks.
To implement in EViews VAR short-run patterns you normally need to specify two matrices, called A and B under VAR / Proc / Estimate Structural Factorizaton. Notice that A (the left hand side matrix) has ones on the diagonal, expressing that @e1 = @e1, etc... Search hlp doc: IDENTIFYING RESTRICTIONS. Maybe it 'd be much more useful for you to start with filling in the matrices manually before you 'd try to write in the restrictions in text form. Good luck.

Hi!

I define the "A" matrix of strucutral restrictions:

1 0 0 0 na na na
na 1 0 0 0 0 na
0 na 1 na 0 0 na
na na na 1 0 0 na
na 0 0 na 1 na 0
na 0 0 na na 1 0
0 0 na 0 na na 1

The B matrix is diagonal (na).

I'm getting the same error message: Hessian of structural VAR likelihood is singular at starting values.

What can I do? Is my restrion matrix wrong?

Thanks!