normal wishart

[gerluc]

hello. when running a normal wishart bayesian var, why is it not possible anymore in EViews11 to set prior for Lambda1?

[EViews Gareth]

It doesn't exist in the Normal-Wishart prior. It is "replaced" by the C hyper parameters.

[gerluc]

what do you mean by "replaced"? In what sense? please elaborate. thank you

[EViews Gareth]

http://eviews.com/help/content/bVAR-Tec ... l#ww265821

[gerluc]

is it because of the H kronecker product Sigma for the prior in beta in equation (44.18)?

[EViews Gareth]

Equations 44.18 through 44.22 have no lambdas. But in previous versions of EViews we abused notation by using lambda instead of C in 44.22.

[gerluc]

C3 is the degrees of freedom parameter. Is that too high to be a prior variance for mu1?

[gerluc]

Or, should i take the inverse of C3 to obtain prior variance of Mu1?

[gerluc]

What does S scale and V scale mean in the dialog box?

[EViews Gareth]

Its a description of what C1 and C2 are.

[gerluc]

Hi Gareth,

Let me come back to this topic. On equation 44.22, page 807 of the Eviews11 manual to obtain the covariance terms of a normal wishart, how come the C1 hyperparameter is not found in the three equations? Should H_inverse of NW = C1*I (identity with dimension k), and not as presented in that same equation?

[EViews Gareth]

Ah, that's a typo.

The first line should be C1, the second line should be C2 and the third should be C3. Apologies.

[EViews Gareth]

hyper.jpg (5.72 KiB) Viewed 16158 times

[gerluc]

Thanks, Gareth. Makes more sense now.