Fit GRF-Backed CATE Models

Usage

grf_cate(
  data,
  modifiers,
  confounders = modifiers,
  treatment,
  outcome,
  ...,
  method = NULL,
  cross_validate = NULL,
  mu.hat = NULL,
  pi.hat = NULL,
  m.hat = NULL,
  treatment_level = 1,
  control_level = 0,
  cross_fit = TRUE,
  folds = 10,
  cv_control = NULL,
  ep_targeting_style = "r",
  ep_r_targeting_basis = "v_plus_propensity",
  tune = c("light", "none", "all"),
  grf_params = list()
)

Arguments

data: A data frame or data.table.
modifiers: Effect modifiers.
confounders: Adjustment covariates.
treatment: Treatment column.
outcome: Outcome column.
...: Additional arguments passed to hte_task().
method: Optional GRF CATE method specification.
cross_validate: Optional logical flag for outer learner selection.
mu.hat: Optional n x 2 matrix of nuisance outcome regressions ordered as (control, treatment).
pi.hat: Optional length-n vector of treatment propensities P(A = treatment_level | X).
m.hat: Optional length-n vector of marginal outcome means.
treatment_level: Treated level for the target contrast.
control_level: Control level for the target contrast.
cross_fit: Whether nuisance learners should be cross-fitted.
folds: Number of folds used for nuisance cross-fitting.
cv_control: Optional outer cross-validation control list.
ep_targeting_style: EP targeting variant used when method includes "ep".
ep_r_targeting_basis: First-stage basis construction used for EP-R.
tune: Tuning mode for the final GRF learner(s).
grf_params: Optional named list of GRF arguments passed to the learners.

Value

An hte3_model.