bvars <img src="images/Logo.png" align="right" width="120" />

Overview

bvar is a collection of R routines for estimating Linear and Nonlinear Bayesian Vector Autoregressive models in R. The original R code was based on the Matlab Code by Blake and Mumtaz (2012) and Koop and Koribilis (2009) and has since then undergone several iterations, extensions and updates.

Models and functionalities include:

• VAR Models

  • Linear VARs
  • Regime Switching VARs
  • Threshold VARs
  • Factor-Augmented Models

• Identification of Structural Models

  • Cholesky decomposition
  • Sign Restrictions
  • Zero restrictions

• Applications

  • Impulse-Response Functions
  • Forecast error variance decomposition <not yet implemented>
  • conditional and unconditional forecasting
  • historical decomposition

• Utilities

  • Plotting of Impulse-Response Functions, Forecasts

• Project Homepage

• Python version (WIP)

Installation

To install the package you need the devtools package. If you don't have the devtools package, you can install it with

install.packages("devtools")

Once you have installed the devtools package you can install the bvar package with

library('devtools')
devtools::install_github('joergrieger/bvars')