Apply all mediation estimators across features (internal)

Usage

run_mediation_methods(
  dat,
  n_features = ncol(dat$Y),
  W_valid = NULL,
  n_cores = 1,
  se_method = "delta",
  n_boot = 500
)

Arguments

dat: List returned by generate_toy_data() / run_single_iteration().
n_features: Number of outcome columns to process.
W_valid: Optional: validity-screened W matrix for matrix-bridge PGC.
n_cores: Number of parallel workers. Default 1 (sequential).
se_method: "delta" (default), "bootstrap", or "composite".
n_boot: Number of bootstrap resamples when se_method="bootstrap".

Value

Data frame with columns: feature, method, NDE, NDE_se, NDE_p, NIE, NIE_se, NIE_p. When se_method="composite", also includes alpha_M, alpha_se, beta_M, beta_M_se, var_a, var_b.