BVAR

[olsisthebest]

Why are you saying that the hyperparameters are crazy?

[EViews Gareth]

Because the values are random machine noise.

[olsisthebest]

OK, thank you.
So you will fix the bug and then an update will be available?

Another question: in the first version of the BVAR addin the user must specify 8 scalars (lambda0 to lambda5, mu5 and mu6) as in Brandt and Freeman's 2006 paper.
In this version one cannot define values for lambda2, lambda 4 and lambda5. Why?

[EViews Gareth]

The BVAR add-in and the built in BVAR estimation are completely separate.

[olsisthebest]

Ok, Gareth but why one can only specify 6 hyperparameters instead of 8 as in Sims-Zha NW approach?
Why cann't we define values for lambda2, lambda 4 and lambda5?
Why cann't we set the mean of the AR(1), mu, also?

[EViews Gareth]

Because it was not implemented to allow you to do so.

[olsisthebest]

It is a tautology :)

I am a newbie in this methodologies.
I am just trying to understand the (un)usefulness of the specifications.
Are there any pratical advantages (in terms of forecasting gains) in using the BVAR add-in instead of the built in BVAR estimation (or vice-versa)?

[EViews Gareth]

Only if you're using EViews 7 (which didn't have the built in BVAR)

[olsisthebest]

And if I want to include exogenous variables in the BVAR model? Since the hyperparameters lambda 4 and lambda5 controls for exogenous variables, should I consider the BVAR add-in instead of the built in BVAR estimation?
What are the implications of do not define lambda4 and lambda5?