Batch Process Subset Models for Causal Forests — margot_subset

We compared treatment effects across multiple subsets derived from causal forest models (using the grf package). For each subset, the number of participants, the total sample size, and the corresponding percentage are computed, describing the composition of each targeted subpopulation. In addition, for each subset the plot, interpretation, and transformed table are generated to summarise the causal effect estimates. Note that exploratory comparisons in causal forest analyses should be interpreted with caution, as subsetting targeted subpopulations may yield unstable treatment effect estimates, typically indicated by large standard errors and wide confidence intervals.

Usage

margot_subset_batch(
  model_results,
  X,
  base_defaults = list(),
  title = "",
  label_mapping = NULL,
  subsets,
  debug = FALSE,
  original_df = NULL,
  ...
)

Arguments

model_results

Results from the causal forest model.

X

Covariate matrix.

base_defaults

A list of default plotting options.

title

A character string for the main title of the plots.

label_mapping

A named vector mapping variable names to labels.

subsets

A named list of subset definitions. Each element should be a list containing:

var: A character string specifying the variable to subset on.
value: The value used to define the subset.
operator: (Optional) A comparison operator (default is "==").
subset_condition: (Optional) A pre-computed logical vector defining the subset.
description: A character string describing the subset.
label: (Optional) A user-friendly label for the subset. If missing, the list name is used.

debug

Logical; if TRUE, prints debugging information.

original_df

Optional data frame containing the original (non-transformed) data for back-transformation.

...

Additional arguments to be passed to margot_plot().

Value

A list with the following elements:

results: A list of subset results, each containing the plot, interpretation, and transformed table along with sample statistics.
summary: A data frame summarising the sample size, total sample size, and percentage for each subset.
explanation: A character string providing a comprehensive explanation of the subsetting process and the results for each model.