Plot allocation scores against the budget or over time

Usage

plot_scores_slim(
  slim_df,
  Ks = NULL,
  ts = FALSE,
  ts_dates = NULL,
  origin_time_col_name = NULL,
  model_col_name = NULL,
  target_col_name = NULL,
  show_oracle = FALSE,
  show_raw = FALSE,
  order_at_K = NULL,
  order_at_model = NULL,
  order_at_origin_time = NULL,
  palette = NULL,
  linetypes = NULL,
  ytot = TRUE
)

Arguments

slim_df: a slim scored allocation; see slim().
Ks: budgets to plot. Defaults to NULL, which keeps all of them; note that the bar layout requires a single budget.
ts: plot the score as a time series over origin_time rather than against the budget. Requires one budget per origin time, given in Ks.
ts_dates: restrict a time series plot to these origin times.
origin_time_col_name: name of the column holding an origin time, used to facet.
model_col_name: name of the column holding a model name. For hubverse model output this is "model_id".
target_col_name: name of the column holding target names.
show_oracle: whether to add the oracle's components as an extra model.
show_raw: plot the forecaster's own losses rather than its losses relative to the oracle.
order_at_K, order_at_model, order_at_origin_time: order the targets by their contribution at this budget, model or origin time. By default the targets are ordered over the whole data set.
palette: colours for the models, passed to ggplot2::scale_color_manual().
linetypes: line types for the models, passed to ggplot2::scale_linetype_manual().
ytot: add a dashed vertical line at the total weighted outcome, the budget beyond which the oracle incurs no loss.

Value

A ggplot2::ggplot() object.

Examples

fc <- add_pdqr_funs(
  tibble::tibble(
    target_names = c("a", "b", "c"), dist = "norm",
    mean = c(5, 8, 12), sd = c(1, 2, 3)
  ),
  types = c("p", "q")
)
scored <- alloscore(fc, y = c(4, 9, 11), K = seq(5, 40, by = 5))
plot_scores_slim(slim(scored))