Select a representative impulse-response draw from the posterior.
Usage
representative_irf(object, ...)
# Default S3 method
representative_irf(object, ...)
# S3 method for class 'PosteriorIR'
representative_irf(
object,
method = c("median_target", "most_likely_admissible"),
center = c("median", "mean"),
variables = NULL,
shocks = NULL,
horizons = NULL,
metric = c("l2", "weighted_l2"),
standardize = c("none", "sd"),
probability = 0.9,
...
)
# S3 method for class 'PosteriorBSVAR'
representative_irf(
object,
horizon = NULL,
method = c("median_target", "most_likely_admissible"),
center = c("median", "mean"),
variables = NULL,
shocks = NULL,
horizons = NULL,
metric = c("l2", "weighted_l2"),
standardize = c("none", "sd"),
probability = 0.9,
...
)Arguments
- object
A posterior model object or a
PosteriorIR.- ...
Additional arguments passed to computation methods.
- method
Representative-model selection method.
- center
Posterior summary used as the target for median-target selection.
- variables
Optional subset of response variables.
- shocks
Optional subset of shocks.
- horizons
Optional subset of horizons.
- metric
Distance metric used for median-target selection.
- standardize
Optional standardisation used in distance computation.
- probability
Equal-tailed interval probability used for summaries.
- horizon
Forecast horizon when
objectis a posterior model object.
Value
A list of class RepresentativeIR (inheriting from
RepresentativeResponse) with elements representative_draw
(the selected IRF array), posterior_draws (all IRF draws),
draw_index (integer index of the selected draw), method,
score, target_summary, selection_spec,
probability, and object_type.
Examples
data(us_fiscal_lsuw, package = "bsvars")
spec <- bsvars::specify_bsvar$new(us_fiscal_lsuw, p = 1)
#> The identification is set to the default option of lower-triangular structural matrix.
post <- bsvars::estimate(spec, S = 5, show_progress = FALSE)
rep_irf <- representative_irf(post, horizon = 3)
rep_irf$draw_index
#> [1] 3