- Add historical decompositions and RMSE/LPS, and WAIC for analysis
- Improve the algorithm to draw sign restrictions
- Shocks are now built and checked per variable
- Fix the CRAN NOTE on missing "tangle output"
- Update FRED-datasets to 2023-10 vintage
- Newer FRED-QD files include (unannounced) small naming changes
- Fix bug from last update, where automatic ARIMA-priors where sqrt'd twice
- Thanks to Michael Wolf for pointing this out, and
- many thanks to Nirai Tomass for discovering the bug
- Update FRED-MD and FRED-QD datasets to the 2023-02 vintage
- Fix bug where warnings caused an error during automatic ARIMA-based priors
- Thanks to Martin Feldkircher for pointing this out
- Add DOI in the CITATION file for a new JSS publication
- DOI will be registered after publication on CRAN
- References now use the DOI interface instead of URLs
- Update FRED-QD and FRED-MD datasets to 2021-10
- Fix minor issues with vignette (e.g. fixed dataset, references, etc)
- Add verbosity to ARIMA-based automatic prior settings
- Add hook to simplify use of shared generics with vars
- Add README file to the package with correct URLs
- Add identification via zero and sign restrictions
- Update CITATION with the forthcoming JSS reference
- Prepare for tidy outputs in BVARverse
- Fix minor bugs and typos (internal checks, documentation, and vignette)
- Add fancy vignette with background and demonstrations
- Add new features
- New FRED-MD database and updated FRED-QD (
data("fred_md")) - Transformation helper functions (
fred_transform(),fred_code()) - Add Conditional forecasting (see
?bv_fcast()) - New replacement functions for
irf()andpredict()(irf(x) <- irf(x)) - Provide wrapper for parallelised execution (
par_bvar())
- New FRED-MD database and updated FRED-QD (
- Improve existing features
- Enhance IRF and forecast plotting
areaargument adds polygons for credible intervalst_backallows adding realised values before forecastscolandfillarguments allow changing colours- transparence is applied to sequential lines / polygons
- Improved x-axis labelling
- Regex may be used for
vars,vars_response, andvars_impulse - New
typeforcoef(),fitted(), etc, to retrieve means / quantiles - Add constructors for the prior mean
bargument inbv_mn() - Auto
psinow allows for one order of integration
- Enhance IRF and forecast plotting
- Enhance speed considerably (~2-10 times faster)
- Move IRF and forecasts out of MCMC
- Capitalise upon matrix properties
- Cached and customised multivariate normal drawing
- Optimised FEVD computation
- Fix bugs
- IRF calculation is now ordered properly (please recalculate)
- coda methods are now proper methods (potential issue on Windows)
- Vectorised
scale_hessnow works properly
- Remove deprecated functions and arguments
- Work on documentation and examples
- Update citation information
- Improve upon internal structure
- Unit tests with tinytest for development (skipped on CRAN)
- Outsource additional steps to dedicated functions
- More robustness checks and add verbosity to errors
- Tested extensively on R 4.0.0 and R 3.6.3
- Fix impulse response calculation (stray transpose of coefficients)
- Thanks, Maximilian Böck for helping us track this down
- Add some verbosity to error messages
- Prepare for next major release
- Include messages about future extensions and changes of FEVDs
- Update docs on future construction of sign restrictions
- Update docs on prediction and IRFs in
bvar()
- Add FRED-MD co-author Serena Ng to data contributors
- Clarify exact ToU of the FRED-QD dataset with St. Louis Fed
- Comply with the new modified ODC-BY 1.0 for FRED-QD
- Mention and add license to LICENSE file (linked in DESCRIPTION)
- Add the copyrighted series we are allowed to use
- Mention updates and license in the data documentation
- Fix and improve documentation
- Extend methods further
- Add
summary()methods forbvar,bvar_irfandbvar_fcast - Add
as.mcmcmethod to interface with coda - Support hyperparameters and coefficients in
plot.bvar()anddensity() - Add
companion()method to retrieve the companion matrix - Improve
ylimofplot.bvar()density plots with multiple chains - Add
logLik()method forbvarobjects
- Add
- Fix
bv_psi()errors for modes other than "auto" - Change
fcastandirfdefaults ofbvar()toNULL - Move from MASS to mvtnorm for
dmvnorm()(used inlogLik()) - Be more specific in error messages
- Add coda to suggestions for convergence assessment etc.
- Add files LICENSE, CITATION and NEWS.md
- Improve examples for further test coverage
R CMD check --run-donttest: One warning (deprecated functions)
- Prepare for minor release 0.2.0
- Provide several standard methods for objects generated with BVAR
predict()for ex-post forecasts and updating quantilesirf()/fevd()for ex-post irfs, fevds and updating quantilesfitted(),residuals(),coef()andvcov()density()as shorthand for callingdensity()on hyperparameters
- Add
print()methods for intermediate objects- Includes
bv_minnesota,bv_metropolisand method outputs
- Includes
- Rework plotting
plot.bvar()now supports types, subsets and multiple chains- Deprecate
bv_plot_trace()andbv_plot_density()- Add
density(), including a plot method - Replace by
plot.bvar(), incl. plotting of multiple chains
- Add
- Change plotting of
bvar_fcastandbvar_irf- Deprecate
bv_plot_irf(x)forplot(x$irf)orplot(irf(x)) - Move
conf_bandsargument topredict()/irf()
- Deprecate
- Add plot methods for
bvar_residandbvar_density
- Add
sign_limtobv_irf()to set maximum number of sign restriction draws - Standardise prior construction further
- Align
alphaandlambda, improvepsialignment bv_mn()may be called with a numeric vector foralphaand/orlambda
- Align
- Improve documentation
- Document related functions in joint
- Add details on priors
- Add less concise aliases for
bv_mh()andbv_mn()bv_metropolis()andbv_minnesota()
- Fix bugs related to single confidence bands at 0.5
- Save
fred_qdwith format version 2, lowering R dependency to (>= 3.3.0) - Add vars to suggests for shared methods, import is bypassed
- Try to clarify licensing terms with the Federal Reserve
- Some copyrighted series may have to be removed
- Subset the dataset to only include variables in public domain for now
- Fix addition of prior pdfs to ML
alphaneeds an sd parameterpsinow needs proper shape and scale parameters- Add normalising constant
- Add lines to all density plots (when supplied via ellipsis)
- Add documentation on using
scale_hessas a vector - Add two pre-constructed dummy priors
bv_soc()andbv_sur() - Further split up calculation of marginal likelihood
- Update DESCRIPTION with linked DOI
- Change
\dontrun{}examples to\donttest{} - Fix bounds in
plot_hyper() - Update references with links via DOI
- Add and improve examples for
printandplotmethods R CMD check --as-cran: No errors or warnings, one note (New submission)