BVAR_linear {BGVAR}R Documentation

Bayesian VAR

Description

This is the sub function that estimates the country VAR models. It is directly called from gvar.ssvs.

Usage

BVAR_linear(nr, gW = gW, xglobal = xglobal, cN, p = 4, nsave = 500, nburn = 500, hyper = list(c_tau = 0.01, d_tau = 0.01, e_lambda = 1.5, d_lambda = 1,  prmean = 0, a_start = 0.6, sample_A = TRUE), prior = "NG", thin = 0.1, sv = TRUE, trend = FALSE)

Arguments

nr

Is a selector referring to the number of the country model to be estimated (same order as in cN below).

gW

Is a list object of length N containing the country-specific weight matrices.

xglobal

is a matrix object of dimension T times N (T # of observations, K # of variables in the system).

cN

Character vector of the country names.

p

Number of lags used (the same for domestic, exogenous and weakly exogenous variables).

nsave

Number of draws saved.

nburn

Number of burn-ins.

hyper

Is a list object that defines the hyperparameters of the chosen prior. See bgvar for further details.

prior

Either "SSVS", "SIMS" or "NG". See the description for gvar.ssvs.

thin

Is a thinning interval which grabs save_thin percent of your posterior output. As a rule of thumb, workspaces get large if save_thin*saves>500.

sv

Set to TRUE implying use of stochastic volatility.

trend

Set to TRUE if a trend should be included. Currently de-funct.

Author(s)

Florian Huber

References

Part of the algorithm is based on Koop and Korobilis.

See Also

See also bgvar.


[Package BGVAR version 1.1.3 Index]