Batch process policy tree interpretations — margot_interpret_policy

Usage

margot_interpret_policy_batch(
  models,
  model_names = NULL,
  max_depth = 2L,
  save_path = NULL,
  prefix = NULL,
  include_timestamp = FALSE,
  report_policy_value = c("none", "treat_all", "control_all", "both", "treated_only"),
  policy_value_R = 499L,
  policy_value_seed = 42L,
  policy_value_ci_level = 0.95,
  brief = FALSE,
  brief_save_to = NULL,
  return_as_list = FALSE,
  ...
)

Arguments

models

A list containing the results from multi-arm causal forest models.

model_names

A character vector of model names to interpret. If NULL, all models are processed.

max_depth

Integer, 1 or 2; which saved policy tree to interpret (default 2).

save_path

The path where the combined interpretation will be saved. If NULL, nothing is saved.

prefix

An optional prefix for the filename.

include_timestamp

Logical; whether to include a timestamp in the filename (if desired).

report_policy_value

Character: one of "none" (default), "treat_all", "control_all", or "both". If not "none", each model interpretation will include a one-line policy value summary with 95

policy_value_RInteger >= 199; bootstrap replicates (default 499).

policy_value_seedInteger or NULL; RNG seed (default 42).

policy_value_ci_levelNumeric confidence level (default 0.95).

briefLogical; if TRUE, prepend a compact treated-only summary for each model (coverage treated and average uplift among treated) and optionally save it.

brief_save_toOptional path to save the brief treated-only summary as text.

...Additional arguments to pass to margot_interpret_policy_tree(), including include_conditional_means (default TRUE), use_math_notation (default FALSE), output_format ("bullet" or "prose"), original_df, label_mapping, and policy value options.

A single character string containing the combined markdown output. This function now accepts a vector of model names to process and produces a single combined output. The common description is printed once at the top, followed by each model's specific findings. You can now control whether to interpret the depth-1 or depth-2 tree via the `max_depth` argument.