EViews.com

Hi,

I am creating a 6 variable SVAR and wanted to use Cholesky decomposition to structure it.

For my A matrix I have made all upper triangular entries 0 and the leading diagonal entries as 1, with all remaining entries as NA.

However, I do not understand what to enter for my B matrix.

When I left all of the entries as NA and entered, a message arose saying that B is singular at starting values which was not allowed.

However, when I structured B in the same way as A, a message popped up saying that the Hessian of SVAR is singular at starting values which is not allowed.

If someone could please explain what I need to input as my B matrix, then I would be most grateful.

Hello,

I'm a little bit confused by your question, so let me propose an alternative to see if that's what you actually want. If you want the Cholesky decomposition of the residuals of the underlying VAR, you can compute that directly, e.g., "matrix chol_decomp = @cholesky(myvar.@residcov)", there is no need to use an SVAR. While you could replicate a Cholesky decomposition with an SVAR, there is no benefit in doing so. SVARs are most useful when you want other types of decompositions.