Summarises response half-lives from posterior draws — half_life

Transforms posterior draws of impulse responses or cumulative dynamic multipliers into summaries of the first horizon at which a response falls to a chosen fraction of its initial or peak level.

Usage

half_life_response(object, ...)

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

# S3 method for class 'PosteriorIR'
half_life_response(
  object,
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  ...
)

# S3 method for class 'PosteriorBSVAR'
half_life_response(
  object,
  horizon = NULL,
  type = c("irf", "cdm"),
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARMIX'
half_life_response(
  object,
  horizon = NULL,
  type = c("irf", "cdm"),
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARMSH'
half_life_response(
  object,
  horizon = NULL,
  type = c("irf", "cdm"),
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARSV'
half_life_response(
  object,
  horizon = NULL,
  type = c("irf", "cdm"),
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVART'
half_life_response(
  object,
  horizon = NULL,
  type = c("irf", "cdm"),
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorBSVARSIGN'
half_life_response(
  object,
  horizon = NULL,
  type = c("irf", "cdm"),
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  scale_by = c("none", "shock_sd"),
  scale_var = NULL,
  ...
)

# S3 method for class 'PosteriorCDM'
half_life_response(
  object,
  variables = NULL,
  shocks = NULL,
  variable = NULL,
  shock = NULL,
  fraction = 0.5,
  baseline = c("peak", "initial"),
  absolute = TRUE,
  probability = 0.9,
  model = "model1",
  ...
)

Arguments

object: A posterior model object, PosteriorIR, or PosteriorCDM.
...: Additional arguments passed to computation methods.
variables: Response variables to include (character or integer vector).
shocks: Structural shocks to include (character or integer vector).
variable: Deprecated. Use variables instead.
shock: Deprecated. Use shocks instead.
fraction: Fraction of the reference level used to define the half-life.
baseline: Reference level: "peak" uses the largest response over the available horizons, "initial" uses the horizon-0 response.
absolute: If TRUE, search for the largest absolute response.
probability: Probability mass of the equal-tailed credible interval.
model: Label identifying the model specification.
horizon: Maximum horizon used when object is a posterior model object.
type: Response type for posterior model objects: "irf" or "cdm".
scale_by: Optional scaling mode for CDMs.
scale_var: Optional scaling variable specification.

Value

A bsvar_post_tbl with columns model, object_type, variable, shock, fraction, baseline, mean_half_life, median_half_life, sd_half_life, lower_half_life, upper_half_life, and reached_prob.

Details

reached_prob is the fraction of stored draws with a defined half-life within the available horizon. A draw has no defined half-life when its reference response is nonpositive or it never reaches the target. Timing summaries condition on draws with a defined half-life; they are NA when none qualify. The standard deviation is NA with one qualifying draw. Missing or nonfinite input responses are rejected.

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)

hl <- half_life_response(post, horizon = 3)
print(hl)
#> # A tibble: 9 × 22
#>   model  object_type   variable shock fraction baseline mean_half_life
#>   <chr>  <chr>         <chr>    <chr>    <dbl> <chr>             <dbl>
#> 1 model1 half_life_irf ttr      ttr        0.5 peak                 NA
#> 2 model1 half_life_irf ttr      gs         0.5 peak                 NA
#> 3 model1 half_life_irf ttr      gdp        0.5 peak                  1
#> 4 model1 half_life_irf gs       ttr        0.5 peak                 NA
#> 5 model1 half_life_irf gs       gs         0.5 peak                 NA
#> 6 model1 half_life_irf gs       gdp        0.5 peak                 NA
#> 7 model1 half_life_irf gdp      ttr        0.5 peak                 NA
#> 8 model1 half_life_irf gdp      gs         0.5 peak                 NA
#> 9 model1 half_life_irf gdp      gdp        0.5 peak                 NA
#> # ℹ 15 more variables: median_half_life <dbl>, sd_half_life <dbl>,
#> #   lower_half_life <dbl>, upper_half_life <dbl>, reached_prob <dbl>,
#> #   analysis_type <chr>, interval_probability <dbl>, interval_method <chr>,
#> #   scale_by <chr>, scale_var <chr>, scale_factor <dbl>, units <chr>,
#> #   source_package <chr>, source_version <chr>, n_draws <int>