A Bayesian vector autoregression (VAR) model assumes a prior probability distribution on all model coefficients (AR coefficient matrices, model constant vector, linear time trend vector, and exogenous regression coefficient matrix) and the innovations covariance matrix. When combined with data to form a posterior distribution, this framework can lead to a more flexible model and intuitive inferences.

To start a Bayesian VAR analysis, create the prior model object that best describes your prior assumptions on the joint distribution of the coefficients and innovations covariance matrix. bayesvarm creates Bayesian VAR models with a Minnesota prior regularization structure. Then, using the prior model and data, estimate characteristics of the posterior distributions, simulate from the posterior distributions, or forecast responses using the predictive posterior distribution.