NEWS
bvartools 0.2.4 (2024-01-08)
- Using an updated version of
Rcpp to address an issue with Rcpp::stop.
stochvol_ocsn2007 can handle multi-column input.stochvol_ksc1998 can handle multi-column input.- Added
post_gamma_state_variance for posterior simulation of constant error variances of the state equation.
- Added
post_normal_covar_tvp for posterior simulation of time varying, lower triangular covariance matrices.
- Added
post_normal_covar_const for posterior simulation of constant, lower triangular covariance matrices.
bvartools 0.2.3 (2023-08-30)
- Fixed alias issue resulting from use of
roxygen2.
- Made
kalman_dk callable from C++.
- Stochastic volatility algorithms allow to set the offsetting constant manually.
- Changed
stoch_vol to a wrapper for stochvol_ksc1998.
- Added stochastic volatility algorithm of Kim et al. (1998) in a separate function
stochvol_ksc1998.
- Added stochastic volatility algorithm of Omori et al. (2007) in function
stochvol_ocsn2007.
- Fixed bug with detection of deterministic terms in
bvar.
- Implemented recursive iterations for forecasts in C++.
- Replaced erroneous
| in C++ sampling functions by ||.
bvartools 0.2.2 (2023-06-12)
- Addressed CRAN NOTE on CITATION file
- Addressed the CRAN NOTE "Specified C++11: please drop specification unless essential" by dropping the specification from "src/Makevars"
- Improved the treatment of
bvar and bvec objects if Gibbs sampler fails.
- Fix erroneous SUR-matrix generation for VEC models with r = 0 in
.bvecalg.
- Fix bug in
.bvecalg and .bvectvpalg with the storing of posterior draws of beta.
- Fix bug of
predict.bvar, which could not handle only VARX models with contemporaneous exogenous variables only.
- Model plot functions support boxplots.
- Fix typos in documentation.
bvartools 0.2.1 (2022-01-22)
- Added functionality for the simulation of models with time varying parameters, both for VAR and VEC models.
- Added functionality for the simulation of models with stochastic volatility, both for VAR and VEC models.
- Added a plot function for classes
bvar and bvec for visual inspection of posterior draws.
- Changed the generation of the output object in the Gibbs sampler functions
bvaralg and bvecalg to make them more stable for especially large output.
- Changed
draw_posterior to a generic function and added the corresponding methods for BVAR, BVEC and DFM input.
- Changed
irf and fevd to generic functions.
- Corrected typos in documentation.
thin_posterior methods were renamed to thin and are now methods of coda::thin.- Function
irf allows to specify the size of a shock.
- Fixed a bug in
ssvs_prior concerning BVEC models.
- Fixed a bug with the prior in the BVEC algorithm.
bvartools 0.2.0 (2021-04-25)
- Changed
thin_posterior to a generic function and added methods for BVAR, BVEC and dynamic factor model input.
- Changed
add_prior to a generic function and added methods for BVAR, BVEC and dynamic factor model input.
- Added funcionality to estimate dynamic factor models (DFM).
predict requires to specify an object of class ts as input for argument exogen.- Additioal argument checks for
add_priors methods.
- Updated documentation in
minnesota_prior and for add_prior methods.
- Using \doi instead of \url in documentation
bvartools 0.1.0 (2020-09-18)
- Omitted package
Matrix from "Imports"" in DESCRIPTION, which caused a note in version 0.0.3.
- Added function
bvarpost for posterior simulation of BVAR models.
- Added function
bvecpost for posterior simulation of BVEC models.
- Added function
draw_posterior for estimation of multiple models.
- Fixed erroneous calculation of structural forecast error variance decompositions.
- More specification checks and increased robustness against erroneous model specificaions.
- Function
fevd calculates FEVDs based on means of posterior draws of FEVDs and not based on the means of the coefficient draws.
- Function
bvar and summary.bvar can deal with inclusion parameters.
- Added funtion
add_priors for easier construction of prior matrices for multiple models.
gen_var and gen_vec can produce multiple models.- Changed all argument names of
predict.bvar to lower cases.
bvartools 0.0.3 (2020-07-23)
- Changed all argument names of
post_normal, post_normal_sur, post_coint_kls and post_coint_kls_sur to lower case letters.
- Replaced output element in function
ssvs from V_i to v_i.
- Refined function
minnesota_prior and added additional functionaliy.
- Fixed error message when creating seasonal dummies with
gen_var and gen_vec.
- New data set
us_macrodata.
- Added additional checks in
gen_vec.
- Added functions
inclusion_prior for the calculation of inclusion probability priors as used by bvs and ssvs.
- Added
summary functions.
- Fixed conversion and collection of exogenous regressors in
bvec_to_bvar.
- Fixed detection of deterministic terms in
bvec_to_bvar.
- Updated documentation in
kalman_dk.
irf contains a new argument keep_draws.- Additional checks in
post_normal, post_normal_sur, post_coint_kls and post_coint_kls_sur.
- Adapt vignette
bvec.
- Added
loglik_normal for the calculation of a multivariate normal log-likelihood.
bvartools 0.0.2 (2019-08-20)
- Updated vignette
ssvs after the introduction of function ssvs_prior.
- Added
ssvs_prior for the calculation of prior matrices for the SSVS algorithm.
- Added
minnesota_prior for the calculation of the Minnesota prior.
- Use unsigned integers for indices in Cpp code to address warnings during installation.
- Better error handling in
irf.
- In
post_coint_kls_sur the prior matrix g_i can be time varying.
bvar and predict also work only with deterministic terms, i.e. p can be zero.- Use SVD to obtain a draw of beta in
post_coint_kls and post_coint_kls_sur.
predict allows for p = 1.- Add legend to
plot.bvarfevd.
bvartools 0.0.1 (2019-06-11)