The RATS Software Forum

There's nothing wrong with that. Isn't that just telling you that the sign of the impact response is unclear?

Sir

The mid-line which represents the impulse response has expected sign (refer figure1). However, error bands are very spread in period 0. That is why I was wondering whether such impulse function is acceptable. Please suggest.

Thanks and Regards

To repeat. There's nothing wrong with that. Isn't that just telling you that the sign of the impact response is unclear?

There is no difference between this and a coefficient in a regression having a confidence band that covers zero.

Hi Sir

Sorry to disturb you again.

Actually I have one small query which I am not able to clear.

Does ordering of variables matters in Structural VAR? I am using identification restrictions suggested by Sims (1986) and Bernanke (1986).

I know in VAR, results are very sensitive to ordering of Variables but I have confusion in case of structural VAR.

Please clarify.

Thanking you

Regards

sanjeev wrote:Hi Sir

Sorry to disturb you again.

Actually I have one small query which I am not able to clear.

Does ordering of variables matters in Structural VAR? I am using identification restrictions suggested by Sims (1986) and Bernanke (1986).

I know in VAR, results are very sensitive to ordering of Variables but I have confusion in case of structural VAR.

Please clarify.

Thanking you

Regards

The short answer is no. The long answer is no, as long as you're careful with the structure of the A and B matrices. You already asked about switching rows where you swapped rows but didn't rearrange the columns to match the intended interpretation of the shocks (https://estima.com/forum/viewtopic.php?p=10684#p10684). In the A and B matrices, rows correspond to variables, columns to shocks.