Batch-compute RATEs for each outcome in a margot_causal_forest result

This replaces the legacy internal RATE code and is the only function that actually talks to grf. It flips the CATE vector on the fly when policy is "withhold_best".

Usage

margot_rate_batch(
  model_results,
  policy = c("treat_best", "withhold_best"),
  target = c("AUTOC", "QINI"),
  level = 0.95,
  round_digits = 3,
  model_prefix = "model_",
  verbose = TRUE,
  seed = 12345
)

Arguments

model_results: List returned by margot_causal_forest(), containing results and full_models.
policy: Character; either "treat_best" (default) or "withhold_best".
target: Character; weighting scheme: "AUTOC" (default) or "QINI".
level: Numeric; Wald confidence level (default 0.95).
round_digits: Integer; decimal places to keep (default 3).
model_prefix: Character; common prefix on model names (default "model_").
verbose: Logical; print progress with cli (default TRUE).
seed: Integer; base seed for reproducible RATE computations (default 12345).