Compare estimates from two forecast aggregations — compare_forecast

Aligns a selected stage from two compatible outputs of aggregate_model_forecasts() and reports signed and absolute differences. Unlike a permissive join, this helper errors when either result has missing or duplicate unit–outcome identities.

Usage

compare_forecast_aggregations(
  first,
  second,
  stage = "consensus",
  unit_by,
  outcomes,
  labels = c("first", "second"),
  scale_width = NULL
)

Arguments

first, second: Named aggregation lists to compare.
stage: Aggregation stage: one of "prompt", "model", "family", or "consensus".
unit_by: Character vector identifying a forecast target.
outcomes: Character vector of outcome columns.
labels: Two names for the estimate columns in the result.
scale_width: Optional scale-width specification: a character column present in both selected stage tables, or a data frame with unit_by, optional outcome, and a numeric scale_width column.

Value

A tibble with both estimates, difference (second - first), absolute_difference, and optional scale-normalized differences.