EViews.com

Hi Oliver,

First, you need to specify and estimate each equation by OLS. (for example eq1.ls y1 c y1(-1) y2(-1)...)
Then make the model using all equations you estimated, identify shocks and simulate it to estimate IRF.

I think that the best way to select the length of moving average and delay parameters is try to look at the impulse response functions.
Check it is sensitive to the length of MA and delay parameter.

Thanks your comment. Now it includes an option to save the IRF to matrix.

When I run the add-in, I always get this message in the end: "_COLMAT is not defined.". However, a favar (var object) and a graph with the IRFs are produced and look fine. WHat would this message mean? Is it something to worry about? Don't worry about this message. If you don't have the v...

That means you have the multicollinearity problem. Check the residuals correlation matrix.

Could you post the file?

After the estimation, variables named _facrot1 _facrot2 ... will be created.

I could not understand the questions. I never mentioned I(0) and I(1) factors. The add-in replicates Bernanke Boivin Eliasz (2005).

The favar add-in is updated. The version 2 includes the variance decomposition and R2

The favar add-in is updated. Now you can do the variance decomposition.

It is working. you need the time series structure.
For example:

No. You can constrain more than one variable.
For example:

var01.confcast "logy logcons" "2017q4 2017q4" "0.02 0.03"

It is impossible for the add-inn.

What is your threshold variable?

First, if you need logged variable you must create the logged variable before the VAR model estimation. Then estimate the VAR model with the created series. For example: series logy=log(y) ... var01.bvar(prior=sznw,initcov=full,l0=0.1,l1=0.2,l3=2,mu1=0.4) 1 4 logy loginc logcons logmdcpi logccpi log...