bsvars 3.1
The package is under intensive development, and more functionality
will be provided soon! To see the package ROADMAP towards
the next version.
Have a question, or suggestion, or wanna get in touch? Join the
package DISCUSSION
forum.
- A NEW MODEL! An SVAR with t-distributed structural shocks
facilitating identification through non-normality is now included in the
package with all the necessary functionality #84
- New ways of verifying identification through heteroskedasticity or
non-normality using method
verify_identification()
#84
- Improve coding of
forecast
cpp
function and R methods #89
- Included or updated legend in FEVD and HD plots as requested by @ccoleman9 #85
bsvars 3.0.1
- Fixed the bugs that started coming up in the new tested version of
Armadillo and RcppArmadillo #82 and RcppCore/RcppArmadillo#443
- Corrected the computations of
verify_autoregression
#82
bsvars 3.0
- The package has a logo! And it’s beautiful! #37
- The package includes
summary
methods #1
- The package includes
plot
methods #36
- Method
forecast
allow for conditional forecasting given
provided future trajectories of selected variables #76
- Sparse mixture and Markov-switching models can now have more than 20
regimes #57
- A new, more detailed, package description #62
- The website features the new logo. And includes some new information
#38
- Updates on documentation to accommodate the fact that some generics
and functions from package bsvars will be used in a
broader family of packages, first of which is bsvarSIGNs.
Includes updates on references. #63
- Fixed
compute_fitted_values()
. Now it’s correctly
sampling from the predictive data density. #67
- Fixed some bugs that did not create problems #55
- Got rid of filling by reference in the samplers for the sake of
granting the exported cpp functions usability #56
- Coded
compute_*()
functions as generics and methods #70
- Updated code for forecast error variance decompositions for
heteroskedastic models (qas prompted by @adamwang15) #69
bsvars 2.1.0
Published on 11 December 2023
- Included Bayesian procedure for verifying structural shocks’
heteroskedastiicty equation-by-equation using Savage-Dickey density
ratios #26
- Included Bayesian procedure for verifying joint hypotheses on
autoregressive parameters using Savage-Dickey density ratios #26
- Included the possibility of specifying exogenous variables or
deterministic terms and included the deterministic terms used by
Lütkepohl, Shang, Uzeda, Woźniak (2023) #45
- Updated the data as in Lütkepohl, Shang, Uzeda, Woźniak (2023) #45
- Fixing the compilation problems reported HERE
#48
- The package has its pkgdown website at bsvars.github.io/bsvars/ #38
bsvars 2.0.0
Published on 23 October 2023
- Included Imports from package stochvol
- Posterior computations for:
- impulse responses and forecast error variance decomposition #3,
- structural shocks and historical decompositions #14
- fitted values #17
- conditional standard deviations #16
- regime probabilities for MS and MIX models #18
- Implemented faster samplers based on random number generators from
armadillo via RcppArmadillo #7
- The
estimate_bsvar*
functions now also normalise the
output w.r.t. to a structural matrix with positive elements on the main
diagonal #9
- Changed the order of arguments in the
estimate_bsvar*
functions with posterior
first to facilitate workflows
using the pipe |>
#10
- Include citation info for the package #12
- Corrected sampler for AR parameter of the SV equations #19
- Added samplers from joint predictive densities #15
- A new centred Stochastic Volatility heteroskedastic process is
implemented #22
- Introduced a three-level local-global equation-specific prior
shrinkage hierarchy for the parameters of matrices and #34
- Improved checks for correct specification of arguments
S
and thin
of the estimate
method
as enquired by @mfaragd #33
- Improved the ordinal numerals presentation for thinning in the
progress bar #27
bsvars 1.0.0
Published on 1 September 2022
- repo transferred from GitLab to GitHub
- repository is made public
- version to be premiered on CRAN
bsvars 0.0.2.9000
- Added a new progress bar for the
estimate_bsvar*
functions
- Developed R6 classes for model specification and
posterior outcomes; model specification includes sub-classes for priors,
identifying restrictions, data matrices, and starting values
- Added a complete package documentation
- Written help files
- Developed tests for MCMC reproducibility
- Included sample data
bsvars 0.0.1.9000
- cpp scripts are imported, compile, and give no
Errors, Warnings, or Notes
- R wrappers for the functions are fully
operating
- full documentation describing package and functions’ functionality
[sic!]