Lrnr_hte Class

This class serves as a general template for constructing meta-learners to estimate heterogeneous treatment effects (HTEs).

Format

An R6 class with public methods to initialize the learner, create a regression task, and access the base learner.

Super class

sl3::Lrnr_base -> Lrnr_hte

Methods

Public methods

• Lrnr_hte$new()

• Lrnr_hte$get_pseudo_data()

• Lrnr_hte$check_treatment_type()

• Lrnr_hte$get_modifiers()

• Lrnr_hte$make_metalearner_task()

• Lrnr_hte$clone()

Inherited methods

• sl3::Lrnr_base$assert_trained()
• sl3::Lrnr_base$base_chain()
• sl3::Lrnr_base$base_predict()
• sl3::Lrnr_base$base_train()
• sl3::Lrnr_base$chain()
• sl3::Lrnr_base$custom_chain()
• sl3::Lrnr_base$get_outcome_range()
• sl3::Lrnr_base$get_outcome_type()
• sl3::Lrnr_base$predict()
• sl3::Lrnr_base$predict_fold()
• sl3::Lrnr_base$print()
• sl3::Lrnr_base$process_formula()
• sl3::Lrnr_base$reparameterize()
• sl3::Lrnr_base$retrain()
• sl3::Lrnr_base$sample()
• sl3::Lrnr_base$set_train()
• sl3::Lrnr_base$subset_covariates()
• sl3::Lrnr_base$train()
• sl3::Lrnr_base$train_sublearners()

---

Method new()

Usage

Lrnr_hte$new(
  params,
  base_learner,
  transform_function = NULL,
  pseudo_outcome_type = c("continuous", "binomial", "quasibinomial"),
  pseudo_family = gaussian(),
  ...
)

Arguments

params

A list of parameters for the meta-learning algorithm.

base_learner

A sl3 learner object inheriting from Lrnr_base that specifies the base supervised learning algorithm used by the meta-learner.

transform_function

A function to transform the predictions of the base learner. Default is the identity transform.

pseudo_outcome_type

The outcome type of the pseudo-outcome used by the meta-learner. Options are c("continuous", "binomial", "quasibinomial"). Default is "continuous". For example, the DR-learner, EP-learner, and R-learner of the CATE involve (weighted) least-squares regression using a pseudo-outcome with pseudo_outcome_type = "continuous". The CRATE EP-learner involves performing weighted logistic regression using a pseudo-outcome taking values in 0,1 with pseudo_outcome_type = "quasibinomial".

pseudo_family

A family object specifying the loss function (involving pseudo-weights and pseudo-outcomes) used to fit base_learner in the meta-learner algorithm. Default is gaussian().

---

Method get_pseudo_data()

Usage

Lrnr_hte$get_pseudo_data(hte3_task, ...)

Arguments

hte3_task

A hte3_Task object containing the data and necessary information for heterogeneous treatment effect estimation.

...

Additional arguments in params needed to compute the pseudo-data.

Returns

A list containing attributes pseudo_outcome and pseudo_weights. (Used internally) Check Compatibility with Treatment Type

This method checks whether this meta-learner is compatible with the treatment variable type in hte3_Task.

---

Method check_treatment_type()

Usage

Lrnr_hte$check_treatment_type(hte3_task)

Arguments

hte3_task

A hte3_Task object containing the data and necessary information for heterogeneous treatment effect estimation.

---

Method get_modifiers()

Usage

Lrnr_hte$get_modifiers(hte3_task, return_matrix = FALSE)

---

Method make_metalearner_task()

Usage

Lrnr_hte$make_metalearner_task(hte3_task, train = TRUE)

Arguments

hte3_task

A hte3_Task object containing the data and necessary information for heterogeneous treatment effect estimation.

train

Logical indicating whether to create the task for training or prediction. Default is TRUE. If FALSE then a hte3_Task object for prediction is returned with covariates being the effect modifiers.

Returns

A sl3_Task object containing the pseudo-data of get_pseudo_data and outcome_Type=pseudo_outcome_type.

---

Method clone()

The objects of this class are cloneable with this method.

Usage

Lrnr_hte$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.