Page 2 of 2

Re: Interpreting impulse response functions: Std dev or % ?

Posted: Mon Jul 23, 2018 8:31 am
by tareb
Hi,

Appologies in advance if my question are very basic, however, i would highly appriciate your feedback.

I am estimating a VAR with Choleski approach. My variables are Bond Yields, Interest Rates on Loan, Investments (growth rates) and GDP growth rates. After i run a VAR and interpreting the IRF i want to have instead of STD the shock to be in %.

1. When STD is used, can i say that one STD is equal to 1% of the dependent variable (in my case Bond Yields)?
2. When using SIRF add-in, what should i include in Impulse variable dialogue box? After i try many option, i get "0 is not a valid index for vector-series-coefficient FAC_DIAG18.

Also, if i am interested only in variable relationships, can i ignor alltogeather the interpretation of coefficents - irespective of their definition in levels of log-levels?

Many thanks,
Tareb

Re: Interpreting impulse response functions: Std dev or % ?

Posted: Wed Jul 25, 2018 8:56 pm
by dakila
1. When STD is used, can i say that one STD is equal to 1% of the dependent variable (in my case Bond Yields)?

No.
2. When using SIRF add-in, what should i include in Impulse variable dialogue box?

Just put one variable, for example just put bond yield variable. Multiple variables are not allowed.
Also, if i am interested only in variable relationships, can i ignor alltogeather the interpretation of coefficents - irespective of their definition in levels of log-levels?

Yes.

Re: Interpreting impulse response functions: Std dev or % ?

Posted: Thu Jul 26, 2018 3:57 am
by Berat
Many thanks, Dakila for the reply, highly appreciated.

When STD are used, what is measured on the horizontal axes of the IRF graph? And also, how would one comment the graph in this case - not just the relationship but in terms of the size of the impact?

Thank you very much.

Re: Interpreting impulse response functions: Std dev or % ?

Posted: Thu Jul 26, 2018 4:33 am
by dakila
It depends on your variable unit. For example, if the bond yield is in % then the horizontal axes is measured in %

Re: Interpreting impulse response functions: Std dev or % ?

Posted: Tue Dec 25, 2018 8:43 am
by Berat
Hi Dakila,

I am still struggling with my interpretation of the PVAR model.

With Choleski decomposition, my variable order is Government Bond Yield (in percentage), Loans to Non-Financial Corporations (in percentage), investment growth rates and GDP growth rates.

The IRF derived are with one STD shock, however, for policy recommendation purposes I want to know the size of 1 STD in terms of my variables in the system. So instead of interpreting 1 STD shock in Government bond yield, i would like to know the 1 STD in terms of my variable Government Bond Yield.

I would appreciate so much your help and you time and so sorry for these basic question.

Berat

Re: Interpreting impulse response functions: Std dev or % ?

Posted: Tue Nov 24, 2020 2:13 pm
by Yuliya
Hello,

please clarify:

1. If I want to rescale Cholesky one std dev to 1% by sirf add-in - what should I put into box "Scale factor for IRF"?
2. Is it depends on form of impulse variable - log, diff log or level?
3. Does sirf add-in work for unrestricted VAR?

Many thanks!

Re: Interpreting impulse response functions: Std dev or % ?

Posted: Thu Dec 31, 2020 8:21 am
by tfeldens
Here just to say thanks for these answers. It took out most of my insecurities in interpreting my results.
God bless you all, have a Happy New Year.