Batch process policy tree interpretations
Source:R/margot_interpret_policy_batch.R
margot_interpret_policy_batch.RdThis 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 depth-1 or depth-2 trees via the `max_depth` argument, or supply per-model depth choices via `depths_by_model`.
Usage
margot_interpret_policy_batch(
models,
model_names = NULL,
max_depth = 2L,
depths_by_model = NULL,
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,
policy_value_source = c("compute", "use_coherent"),
coherent_values = NULL,
...
)Arguments
- models
A list containing the results from multi-arm causal forest models.
- model_names
A character vector of model names to interpret. If named, the values should be 1 or 2 (depth assignments) and take precedence over `max_depth` and `depths_by_model`. If NULL, all models are processed.
- max_depth
Integer, 1 or 2; fallback depth used when no per-model depth mapping is provided via `model_names` or `depths_by_model` (default 2).
- depths_by_model
Optional named vector/list mapping models (with or without `model_` prefix) to depth 1 or 2; combined with any depth hints supplied through `model_names`.
- 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% CIs based on bootstrap SEs.
- policy_value_R
Integer >= 199; bootstrap replicates (default 499).
- policy_value_seed
Integer or NULL; RNG seed (default 42).
- policy_value_ci_level
Numeric confidence level (default 0.95).
- brief
Logical; if TRUE, prepend a compact treated-only summary for each model (coverage treated and average uplift among treated) and optionally save it.
- brief_save_to
Optional 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.