by Matthieu Droumaguet, Anders Warne, & Tomasz Woźniak

Summary. A block Metropolis-Hastings algorithm for the Bayesian estimation of the Markov-switching Vector Autoregressive models with restrictions for Granger noncausality is provided, as well as an appropriate estimator for the marginal data density.

Keywords. R, Markov-switching VARs, Block Metropolis-Hastings Sampler, Marginal Data Density

To refer to the code in publications, please, cite the following paper:

Droumaguet, M., Warne, A., Woźniak, T. (2017) Granger Causality and Regime Inference in Markov-Switching VAR Models with Bayesian Methods, Journal of Applied Econometrics, 32(4), pp. 802--818, DOI: 10.1002/jae.2531.

The project's file structure includes:

• BayesianMSVAR.pdf - a document presenting the model, main functions, and their application
• BayesianMSVAR-example.R - a file presenting code application for a simple example
• BayesianMSVAR - a folder containing the functions for the estimation of the considered models
• ReproductionScripts - a folder containing scripts for the reproduction of all the results contained in the JAE paper
• data.csv and data.RData - data used in the paper

To download the code simply click on the download icon on the top of this page

and select the format of the compressed file to be downloaded.

You can also choose to fork the project and clone your copy from the repository. Just follow the steps (requires a GitLab account):

1. On this page: fork the project by clicking the icon on the top of this page (more info)

   

2. On you computer: clone the repository you have just forked to create its local copy that you can work with.

3. Optional: if you find a bug or if you improve the code, please feel free to submit a pull/merge request. (more info)