Parallel flip of CATE estimates with reproducible RNG and memory guard

Usage

margot_flip_forests_parallel(
  model_results,
  flip_outcomes,
  model_prefix = "model_",
  recalc_policy = TRUE,
  parallel_policy = FALSE,
  n_policy_cores = future::availableCores() - 1,
  verbose = TRUE,
  n_cores = future::availableCores() - 1,
  max_size_GB = 2,
  seed = TRUE
)

Arguments

model_results: A list containing the model results from margot_causal_forest().
flip_outcomes: A character vector of outcome variable names for which CATE estimates should be flipped.
model_prefix: A character string indicating the prefix used for model names in the results list. Default is "model_".
recalc_policy: Logical; if TRUE (default) recalculates policy trees for flipped outcomes.
parallel_policy: logical. if TRUE, policy-tree refits are parallelised.
n_policy_cores: integer. number of workers for policy-tree refits.
verbose: Logical indicating whether to display detailed messages during execution. Default is TRUE.
max_size_GB: numeric. globals cap passed to `future.globals.maxSize` (default 2 GB).
seed: `TRUE` for automatic parallel-safe RNG, or an integer for deterministic streams.