[R-pkgs] New R package ``bvarsv''

Krueger, Fabian fabian.krueger83 at gmail.com
Fri Sep 12 10:15:34 CEST 2014


Dear R users,

please let me draw your attention to my new R package bvarsv (on CRAN 
since August 28) which implements the Primiceri (Review of Economic 
Studies, 2005) vector autoregressive model. The model is popular in 
macroeconomic analysis as it allows to model instabilities (e.g. in the 
mean and volatility dynamics) that are often observed in macroeconomic 
time series like inflation.

The package provides functionality for Bayesian analysis of the 
Primiceri model. To the best of my knowledge, it is the only publicly 
available R code to do so. The underlying Markov Chain Monte Carlo 
algorithm is computationally challenging, since each iteration requires 
multiple calls of a recursive Kalman filter type algorithm. Therefore, 
bvarsv relies on C++ code ported to R via Rcpp and RcppArmadillo. The 
clear focus of the package is on forecasting, as opposed to structural 
analysis of economic relationships. The package allows to compute 
posterior predictive distributions, which can then be plotted and 
analyzed using forecast evaluation techniques. More detailed 
documentation and examples is available here:

https://sites.google.com/site/fk83research/code

Your feedback on any aspects of the package, as well as possible 
improvements or extensions, would be highly appreciated.

Thank you and best wishes,
Fabian Krüger



More information about the R-packages mailing list