Object Reference : Object View and Procedure Reference : Var
  
 
bvar
Estimate a Bayesian VAR specification.
Syntax:
var_name.bvar(options) lag_pairs endog_list [@ exog_list]
bvar estimates an Bayesian VAR. You must specify the order of the VAR (using one or more pairs of lag intervals), and then provide a list of series or groups to be used as endogenous variables. You may include exogenous variables such as trends and seasonal dummies in the VAR by including an “@-sign” followed by a list of series or groups. A constant is automatically added to the list of exogenous variables; to estimate a specification without a constant, you should use the option “noconst”.
Options
General options
 
noconst
Do not include a constant in exogenous regressors list.
prior = keyword (default= “lit”)
Set the prior type: “lit” (Litterman/Minnesota), “nw” (Normal-Wishart), “sznw” (Sims-Zha Normal-Wishart prior), “sznf” (Sims-Zha Normal-flat prior), “glp” (Giannone, Lenze, and Primiceri).
initcov = keyword (default = “uni”)
Set the (initial) residual variance-covariance: “full” (full Classical VAR), “uni” (univariate AR), “diag” (diagonal of full classical VAR), “arconst” (univariate AR and a constant)
By default, EViews uses the “initcov=uni” option so that diagonal elements of the prior residual variance-covariance can be obtained from the estimation of a set of univariate AR models.
nodf
Do not degree-of-freedom correct the initial residual covariance.
initexog
Use exogenous variables in initial covariance estimate.
initdummied
Use dummy variables in initial covariance estimate. (Only applicable if dummy variable options are below).
icsmpl = arg
Set the sample used for initial covariance estimate (estimation sample used if omitted).
sumcoef
Use the sum-of-coefficients dummy variable (only applicable if not using Sims-Zha or GLP prior).
initobs
Use the initial-observations dummy variable (only applicable if not using Sims-Zha or GLP prior).
nsumcoef
Do not use sum-of-coefficients dummy variable (only applicable if using Sims-Zha or GLP prior).
ninitobs
Do not use initial-observations dummy variable (only applicable if using Sims-Zha or GLP prior).
l0 = num
Set the residual covariance tightness hyper-parameter.
l1 = num
Set the overall tightness hyper-parameter.
l2 = num
Set the relative cross-variable weight hyper-parameter.
l3 = num
Set the lag decay hyper-parameter.
l4 = num
Set the exogenous variables hyper-parameter.
l5 = num
Set the other exogenous variables hyper-parameter.
mu1 = num
Set the AR(1) coefficient dummies hyper-parameter.
mu5 = num
Set the sum of coefficient dummies hyper-parameter.
mu6 = num
Set the initial observation dummies hyper-parameter.
c1 = num
Set the S scale hyper-parameter.
c2 = num
Set the V scale hyper-parameter.
c3 = num
Set the degrees-of-freedom hyper-parameter.
optl1
Optimize the L1 hyper-parameter.
optl3
Optimize the L3 hyper-parameter.
optmu5
Optimize the MU5 hyper-parameter.
optmu6
Optimize the MU6 hyper-parameter.
optpsi
Optimize the initial covariance estimates.
draws = num
Set the number of MCMC draws (only applicable with “prior=inw”).
seed = num
Set the seed for the MCMC generator (only applicable with “prior=inw”).
burn = num
Set the percentage of MCMC draws to use as a burn-in (only applicable with “prior=inw”).
m=integer
Set maximum number of iterations (only applicable with “prior=glp”).
c=scalar
Set convergence criterion (only applicable with “prior=glp”).
prompt
Force the dialog to appear from within a program.
p
Print basic estimation results.
Examples
var mvar.bvar 1 3 m1 gdp
declares and estimates an Bayesian VAR with a Litterman prior named MVAR with two endogenous variables (M1 and GDP), a constant and 3 lags (1 through 3).
mvar.bvar(noconst) 1 3 ml gdp
estimates the same VAR, but with no constant.
var mvar.bvar(prior=nw, mu1=0.2, C1=0.2) 1 3 m1 gdp
specifies a normal-Wishart with hyper-prior values specified as mu1=0.2, C1=0.2.
var mvar.bvar(prior=inw, c1=0.15,c2=0.15, draws=50000) 1 3 m1 gdp
specifies an independent normal-Wishart prior with both C1 and C2 hyper-parameters set equal to 0.15, and the number of Gibb’s sampler draws set to 50,000.
Var mbvar.bvar(prior=glp, initcov=diag, nsumcoef, ninitobs, optl3, m=5000) 1 3 m1 gdp
Specifies a Giannone-Lenza-Primeceri prior, where only Lambda 3 will be optimized, the initial covariance matrix is estimated as a diagonal VAR, neither initial dummy observation priors, and the number of optimization iterations set to 5,000.
Cross-references
See “Bayesian VAR Models” for details.
See also Var::ls and Var::ec for estimation of ordinary VARs and error correction models.