Page 1 of 1

SVAR

Posted: Fri Apr 20, 2018 7:01 pm
by Azza
Hi
A - am using eviews 10 and am doing SVAR with 5 variables and I want to make sure that am doing the right process:
1 - select VAR
2 - use proc and estimate structural factorization and I impose restriction on B to be diagonal and on A I impose three more restriction on the lower triangular since I only to impose restriction on the contemporaneous restriction and no restriction on lags
3 - for IR I choose structural decomposition
4 - VD also I select structural decomposition

B- how I can change the analytic in IR from 68% error band?
C- how I can do the same process in A by using syntax?

thanks

Re: SVAR

Posted: Mon Apr 23, 2018 8:58 am
by EViews Matt
Hello,

Regarding (B), you'll need to either use the sirf EViews add-in or manually extract the standard error information to reconstruct the graph(s) with the band size you want. This thread discusses both those options.

Regarding (C), the easiest way to get started is the perform those steps manually via the GUI and then look at the Command Capture output. You can use that code as the starting point for an EViews program.

Re: SVAR

Posted: Fri Jul 06, 2018 12:02 pm
by qabil
Hi everyone,

I have a question regarding conditional forecasts from a SVAR. I proceed as follows:

- Run an unrestricted VAR
- Impose long-run restrictions via LR matrix
- Estimate the SVAR
- make model svarmod
- Produce IRs via structural decomposition

The IRs of SVAR are different from UVAR. Fine until now!

However, when I 'exclude/override' some variables to forecast from model (svarmod), the forecasts are exactly the same as those from UVAR.

Do I miss something???

Thanks!