IRF() argument shock renamed to impulse, for better clarity.
If you have shock = in existing calls to IRF(), rename it to impulse = .
forecast(), conditional_forecast(), and IRF() now transform each
posterior draw before summarizing, rather than summarizing and then transforming.
This fixes incorrect prediction/credible intervals and median point estimates/predictions
when using growth_rate_idx argument. Mean point estimates/predictions are unaffected.
Now d_pred is automatically created in fit() to make the
user-experience smoother. However, it is still left as an argument.
Fixed IRF() plotting: specifying only one of response/impulse
(leaving the other NULL) previously ignored the specified index and
plotted all response-impulse combinations. Now the specified index is
correctly held fixed while the other dimension varies.
New steady_state_priors_plot() function for visualizing steady-state priors.
forecast() now overlays posterior steady-state estimates on forecast
plots.
Stan code has been optimized. This includes assuming that all prior covariance matrices are diagonal.
Added a vignette vignette("SteadyStateBVAR-intro")
Function documentations, vignettes, and the README have been revamped quite extensively.
Stan code now uses the new array syntax (@andrjohns, #4).
Small rework of plotting in forecast() and corrected documentation.