Summarize forecast disagreement within units — summarize_forecast

Computes distribution summaries across any contributor type, such as models or prompt variants. Missing estimates are counted explicitly in n_nonmissing and n_missing; with na_rm = FALSE, any missing estimate makes the numeric summaries for that unit missing.

Usage

summarize_forecast_disagreement(
  data,
  unit_by,
  estimate_col = "estimate",
  contributor_col,
  na_rm = TRUE,
  scale_width_col = NULL
)

Arguments

data: A data frame containing contributor-level estimates.
unit_by: Character vector identifying the units to summarize within.
estimate_col: Numeric estimate column.
contributor_col: Contributor or source identifier column.
na_rm: Whether to omit missing estimates from numeric summaries.
scale_width_col: Optional positive numeric column giving the width of the outcome scale within each unit. If supplied, normalized SD, MAD, and range columns are returned.

Value

A tibble with contributor counts, missingness counts, and mean, median, SD, MAD, minimum, maximum, and range.