Confidence Interval Data of VAR model

Questions and discussions on Vector Autoregressions
bok1234
Posts: 86
Joined: Tue Jan 11, 2011 8:14 pm

Confidence Interval Data of VAR model

Unread post by bok1234 »

Dear Mr.Doan

I want to draw ALL graphs which I posted a few days ago, ‘with EXCEL’, not copying RATS graphs.
'ALL graphs' means not only IRF but also confidence intervals of each panel.
To do this, I need whole numbers of not only every IRF (black line) but also every confidence interval (CI) limits (blue lines). In the case of IRFs, they can be got if ‘noprint’ option is deleted in ‘IMPULSES.RPF’. However, CIs cannot be done through this way. I tried one poor action – deleting ‘noprint’ option in ‘montevar.src’. The result was disaster – just the pages of output got exploded. Would you show me the way how to get the data of upper and lower limits of CIs in every IRF graph panel?
TomDoan
Posts: 7814
Joined: Wed Nov 01, 2006 4:36 pm

Re: Confidence Interval Data of VAR model

Unread post by TomDoan »

Replace the @MONTEVAR with the combination of @MCVARDODRAWS (which does the Monte Carlo integration) and @MCPROCESSIRF (which analyzes the draw information and produces the limits). You can then do a COPY instruction to get the information into an Excel file.

@mcvardodraws(model=canmodel,draws=4000)
@mcprocessirf(model=canmodel,lower=lower,upper=upper,center=median,irf=irfs)
copy(format=xls,org=columns) / irfs upper lower
close copy
bok1234
Posts: 86
Joined: Tue Jan 11, 2011 8:14 pm

Re: Confidence Interval Data of VAR model

Unread post by bok1234 »

Dear Mr.Doan,

Yes, 2 procedures replacing @montevar in 'impulses.rpf' produce successfully excel data of IRFs & CI.
Thank you for your speedy answer as usual.
I was out of office for 3 days, so late for testing your code.
bok1234
Posts: 86
Joined: Tue Jan 11, 2011 8:14 pm

Re: Confidence Interval Data of VAR model

Unread post by bok1234 »

Dear Mr.Doan,

In cases of @MONTEVAR, @MCVARDODRAWS, and @MCPROCESSIRF, every shock is 1 standard deviation of each variable?
TomDoan
Posts: 7814
Joined: Wed Nov 01, 2006 4:36 pm

Re: Confidence Interval Data of VAR model

Unread post by TomDoan »

@MCPROCESSIRF doesn't do anything but process output from something else that generates responses.

By default, @MONTEVAR and @MCVARDODRAWS do unit shocks in the Cholesky factor, which means one standard deviation in each of the orthogonalized shocks, which is not the same as one standard deviation in each variable, except for the first variable in the system. If you want something else, you can use the FFUNCTION option.
Thao NGUYEN
Posts: 1
Joined: Fri Oct 07, 2016 8:46 am

Re: Confidence Interval Data of VAR model

Unread post by Thao NGUYEN »

TomDoan wrote:Replace the @MONTEVAR with the combination of @MCVARDODRAWS (which does the Monte Carlo integration) and @MCPROCESSIRF (which analyzes the draw information and produces the limits). You can then do a COPY instruction to get the information into an Excel file.

@mcvardodraws(model=canmodel,draws=4000)
@mcprocessirf(model=canmodel,lower=lower,upper=upper,center=median,irf=irfs)
copy(format=xls,org=columns) / irfs upper lower
close copy
Hello,
I'm trying to estimate a panel VAR with shrinkage estimators. How can I draw the confidence bands for the impulses responses accumulated in a panel VAR? Could I use @MCVARDODRAWS?
Many thanks.
Nguyen.
TomDoan
Posts: 7814
Joined: Wed Nov 01, 2006 4:36 pm

Re: Confidence Interval Data of VAR model

Unread post by TomDoan »

That was asked and answered in another thread.

https://estima.com/forum/viewtopic.php? ... =15#p15486
Post Reply