i_mp in BVAR

[BGV]

Hi,

To estimate bayesian VAR in Eviews 8 under the Minnesota prior one should specify the value of hyperparameter Mu1 for the prior mean of the VAR parameters. In the user guide the following is stated : theta_0=Mu_1*i_mp, but there is nothing said about the definition of i_mp.

Is the i_mp an identity matrix with mxp dimension? Thus, to shrink VAR to random walk, the Mu_1 should be set at 1, shouldn't it? Is BVAR estimated with OLS? Is there any way to choose optimally the order of BVAR in EViews?

Thank you in advance!

[EViews Gareth]

i_mp is defined in the manual as an mp-element unit vector.

[BGV]

Hi,

Thank you for your reply!

Just one more question. If I set Mu_1=0 (by default), then are the priors for the first lags of the variables in BVAR set to 0 or are they kept at 1 according to the idea of Minnesota prior?

Please, help, this is somewhat confusing.
Thank you again!

[EViews Gareth]

Zero.