Page 1 of 1

Standard errors for structural VAR

Posted: Mon May 05, 2014 6:28 am
by nvellodi
Dear all,

I am estimating a structural VAR (SVAR) in Eviews 7, and want to plot the structural impulse response functions with error bands. Here is the relevant portion of my code:

Code: Select all

var gdpvar.ls 1 2 dlogy zic gdpvar.append(svar) @lr1(@u1)=0 freeze(out1) gdpvar.svar(rtype=text) gdpvar.impulse(40, m, imp=struct, se=mc, smat=stnderr, matbys=rsp_acc)
So I am using the Monte Carlo method for generating the standard errors. I realize that to plot the irfs with error bands, I will need to do it manually, i.e. store the standard error values and then create series that equal the irf plus/minus the stadard errors. However, I can't work out how to store the standard errors generated as output. I thought that the smat option would do it, but this just seems to store the actual impulse response series. Any ideas?

Many thanks

Re: Standard errors for structural VAR

Posted: Sat Jan 02, 2016 4:28 pm
by NicolasR
Hi,

I am estimating some impulse response functions with an SVAR model with short-run restriccions, I want to obtain the confidence intervals for the responses with monte carlo, but when I select the option the interval does not apper. Is it possible to obtain the confidence intervals with the model i use?

Thanks!

Re: Standard errors for structural VAR

Posted: Sun Jan 03, 2016 7:21 am
by trubador
Unfortunately, EViews does not produce error bands for SVAR, VEC and BVAR at the moment. Doing so by programming would also be difficult, since multivariate random sampling procedures (like inverse wishart) are not currently available.

Re: Standard errors for structural VAR

Posted: Wed Apr 05, 2017 2:15 am
by nasa
Hi,

I want to create IRFs with monte carlo standard errors for structural decomposition of AB model. So,please how I can do it in Eviews or any other solution in other statistical packages? Any updates!

Best Regards,

Re: Standard errors for structural VAR

Posted: Wed Apr 05, 2017 8:46 am
by EViews Matt
Hello,

I'm afraid Monte Carlo SEs are still not supported for SVAR IRFs.