SVAR Error Messages

[JohnLWright]

Good Morning,

I am trying to estimate a 5 variable Structure VAR (looking at the impact of monetary policy shocks on commodity prices). When I estimate the structural factorisation I get the following error message "Hessian matrix is near singular at final iteration parameter values.

When I go in a set my B matrix diagonal inputs to 1 instead of N/A I am getting the error message "The B matrix is fixed and the structural innovation variances are not estimated. Are you sure you want to proceed".

I was wonder if others had encountered the error message as well. For reference I am trying replicate the results from the following study

https://www.ecb.europa.eu/pub/pdf/scpwp ... 15943d6edf

Any guidance or support would be appreciated

John

[dakila]

Did you get the data?

[JohnLWright]

Hi Dakila,

Following on from the paper, I aggregated the data from various sources (FFR, M2, CPI and IP from FRED + commodity price indices from the world bank).

John W

[EViews Matt]

Hello,

The "Hessian matrix is near singular..." error occurs when the estimation hasn't converged properly, usually due to identification issues in the model. Sometimes this can be worked around by starting the estimation with different initial values. You might try repeatedly re-estimating the SVAR with random initial values in the hopes of happening upon values that permit convergence.

[EViews Gareth]

Following on from Matt's suggestion - you might want to try starting values that are close to those given in the study you are trying to replicate.

[JohnLWright]

Thanks Matt and Gareth. I will attempt your suggestion.

Regards

John