bayesianVARs (development version)

added the possibility of estimating a VAR(0) (VAR with zero lags) and of specifying 0 factors for the factor error specification.
improved error handling; among other things replaced Rf_error() with Rcpp::stop().

bayesianVARs 0.1.8

fixed CRAN check (memory access) issues. Thanks to Brian Ripley for pointing them out.

Welcome Stefan Haan as contributor to the package!
bayesianVARs now features impulse response functions via irf(). Thanks to Stefan for taking care of this!
bvar() now formally can estimate 'huge' VARs in combination with a factor specification on the errors by setting 'expert_huge=TRUE'.
Underlying 'cpp' functions of bvar(), predict() and vcov() now return Rcpp objects instead of arma objects. Efficiency gains only measurable in huge dimensions!
new additional unit tests (bayesianVARs/test/testthat)
internal restructuring of specify_prior_phi() and specify_prior_sigma(), for users of bayesianVARs nothing changes
Updated references in DESCRIPTION, vignette and function documentations.

Bug fix which makes estimation possible of a VAR with factor stochastic volatility structure on the errors with the restriction that the loadings matrix has zeros above the diagonal. Thanks to Stefan Haan for reporting the bug.

For consistency with other functions, from now on prior_intercept in bvar() specifies standard deviations instead of variances.
Bug fix concerning 'additional check' valgrind. Seems to pass the check now. Thanks to Brian Ripley for reporting the bug.

bugfix concerning VAR with factor structure on errors with homoscedastic factors.

Added minimum version to factorstochvol in the Imports field of the DESCRIPTION file in order to avoid unnecessary building errors. Thanks to Sergey Fedorov for pointing this out.
vcov.bayesianVARs_bvar method now can be specified for specific time-points.
bugfix in cpp function which constructs variance-covariance matrices. If a Cholesky structure for the errors had been specified, exported functions such as vcov, predict and fitted were affected.