Compares posterior forecasts across model specifications

Summarises draws from the predictive distributions of two or more posterior model objects in a common table identified by model specification.

Usage

compare_forecast(..., horizon = NULL, probability = 0.9, draws = FALSE)

Arguments

...

Posterior model objects or a named list of model objects.

horizon

Forecast horizon. If NULL (default), resolves to 20 periods.

probability

Probability mass of the equal-tailed credible interval.

draws

If TRUE, report individual posterior draws rather than posterior summaries.

Value

A bsvar_post_tbl combining results across models, with a model column identifying each model specification.

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.
post1 <- bsvars::estimate(spec, S = 5, show_progress = FALSE)
post2 <- bsvars::estimate(spec, S = 5, show_progress = FALSE)

comp <- compare_forecast(m1 = post1, m2 = post2, horizon = 3)
head(comp)
#> # A tibble: 6 × 19
#>   model object_type variable horizon   mean median    sd lower upper
#>   <chr> <chr>       <chr>      <dbl>  <dbl>  <dbl> <dbl> <dbl> <dbl>
#> 1 m1    forecast    ttr            1  -9.25  -9.09 0.607 -10.1 -8.81
#> 2 m1    forecast    ttr            2  -9.92  -8.98 2.47  -13.3 -8.56
#> 3 m1    forecast    ttr            3 -10.9   -9.16 4.10  -16.5 -8.92
#> 4 m1    forecast    gs             1  -9.76  -9.83 0.256 -10.1 -9.51
#> 5 m1    forecast    gs             2  -9.72  -9.68 0.449 -10.2 -9.20
#> 6 m1    forecast    gs             3 -10.5   -9.53 2.36  -13.7 -9.15
#> # ℹ 10 more variables: 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>