Computes posterior draws of cumulative dynamic multipliers

Each draw from the posterior distribution of the impulse responses is transformed into a draw from the posterior distribution of cumulative dynamic multipliers by summing responses from horizon zero onwards.

Usage

cdm(object, ...)

# Default S3 method
cdm(object, ...)

# S3 method for class 'PosteriorBSVAR'
cdm(
  object,
  horizon = NULL,
  probability = 0.9,
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARMIX'
cdm(
  object,
  horizon = NULL,
  probability = 0.9,
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARMSH'
cdm(
  object,
  horizon = NULL,
  probability = 0.9,
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARSV'
cdm(
  object,
  horizon = NULL,
  probability = 0.9,
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVART'
cdm(
  object,
  horizon = NULL,
  probability = 0.9,
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARSIGN'
cdm(
  object,
  horizon = NULL,
  probability = 0.9,
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

Arguments

object: A posterior model object.
...: Additional arguments passed to methods.
horizon: Forecast horizon.
probability: Probability mass of the equal-tailed credible interval.
scale_by: Either "none" or "shock_sd". The latter divides each cumulative response by the selected observed variable's sample standard deviation, computed from the fitted sample.
scale_var: Observed variable name or index used for shock_sd scaling; a scalar applies to every shock, or supply one entry per shock. With NULL, variables are matched to shocks by position.

Value

A 4-dimensional array of class PosteriorCDM with dimensions [variables x shocks x (horizon + 1) x draws].

Interpretation

The cumulative response at horizon H sums draw-level impulse responses from zero through H (inclusive). Units follow the input transformations and fitted structural-shock normalization. For log-level data this is a cumulative log response, not a dollar-per-dollar fiscal multiplier. shock_sd scaling does not form an output-to-spending response ratio.

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)

cdm_draws <- cdm(post, horizon = 3)
dim(cdm_draws)
#> [1] 3 3 4 5