Movatterモバイル変換

Title:	Tidy Estimation of Heterogeneous Treatment Effects
Version:	1.0.4
Description:	Estimates heterogeneous treatment effects using tidy semantics on experimental or observational data. Methods are based on the doubly-robust learner of Kennedy (2023) <doi:10.1214/23-EJS2157>. You provide a simple recipe for what machine learning algorithms to use in estimating the nuisance functions and 'tidyhte' will take care of cross-validation, estimation, model selection, diagnostics and construction of relevant quantities of interest about the variability of treatment effects.
URL:	https://github.com/ddimmery/tidyhtehttps://ddimmery.github.io/tidyhte/index.html
BugReports:	https://github.com/ddimmery/tidyhte/issues
License:	MIT + file LICENSE
Encoding:	UTF-8
RoxygenNote:	7.3.2
Suggests:	covr, devtools, estimatr, ggplot2, glmnet, knitr, mockr,nprobust, palmerpenguins, quadprog, quickblock, rmarkdown,testthat (≥ 3.0.0), vimp, WeightedROC
Config/testthat/edition:	3
Imports:	checkmate, dplyr, lifecycle, magrittr, progress, purrr, R6,rlang, SuperLearner, tibble
VignetteBuilder:	knitr
NeedsCompilation:	no
Packaged:	2025-07-29 13:55:01 UTC; drewd
Author:	Drew Dimmery [aut, cre, cph]
Maintainer:	Drew Dimmery <cran@ddimmery.com>
Repository:	CRAN
Date/Publication:	2025-07-29 19:20:02 UTC

tidyhte: Tidy Estimation of Heterogeneous Treatment Effects

Description

Estimates heterogeneous treatment effects using tidy semantics on experimental or observational data. Methods are based on the doubly-robust learner of Kennedy (2023)doi:10.1214/23-EJS2157. You provide a simple recipe for what machine learning algorithms to use in estimating the nuisance functions and 'tidyhte' will take care of cross-validation, estimation, model selection, diagnostics and construction of relevant quantities of interest about the variability of treatment effects.

Details

The best place to get started withtidyhte isvignette("experimental_analysis") whichwalks through a full analysis of HTE on simulated data, orvignette("methodological_details")which gets into more of the details underlying the method.

Author(s)

Maintainer: Drew Dimmerycran@ddimmery.com (ORCID) [copyright holder]

References

Kennedy, E. H. (2020). Towards optimal doubly robust estimation of heterogeneouscausal effects.arXiv preprint arXiv:2004.14497.

See Also

The core public-facing functions aremake_splits,produce_plugin_estimates,construct_pseudo_outcomes andestimate_QoI. Configuration is accomplished throughHTE_cfgin addition to a variety of related classes (seebasic_config).

Configuration of a Constant Estimator

Description

Constant_cfg is a configuration class for estimating a constant model.That is, the model is a simple, one-parameter mean model.

Super class

tidyhte::Model_cfg ->Constant_cfg

Public fields

Methods

Public methods

• Constant_cfg$new()

• Constant_cfg$clone()

Methodnew()

Create a newConstant_cfg object.

Usage

Constant_cfg$new()

Returns

A newConstant_cfg object.

Examples

Constant_cfg$new()

Methodclone()

The objects of this class are cloneable with this method.

Usage

Constant_cfg$clone(deep = FALSE)

Arguments

Examples

## ------------------------------------------------## Method `Constant_cfg$new`## ------------------------------------------------Constant_cfg$new()

Configuration of Model Diagnostics

Description

Diagnostics_cfg is a configuration class for estimating a variety ofdiagnostics for the models trained in the course of HTE estimation.

Public fields

Methods

Public methods

• Diagnostics_cfg$new()

• Diagnostics_cfg$add()

• Diagnostics_cfg$clone()

Methodnew()

Create a newDiagnostics_cfg object with specified diagnostics to estimate.

Usage

Diagnostics_cfg$new(ps = NULL, outcome = NULL, effect = NULL, params = NULL)

Arguments

Returns

A newDiagnostics_cfg object.

Examples

Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE", "RROC"),   ps = c("SL_risk", "SL_coefs", "AUC"))

Methodadd()

Add diagnostics to theDiagnostics_cfg object.

Usage

Diagnostics_cfg$add(ps = NULL, outcome = NULL, effect = NULL)

Arguments

Returns

An updatedDiagnostics_cfg object.

Examples

cfg <- Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE", "RROC"),   ps = c("SL_risk", "SL_coefs"))cfg <- cfg$add(ps = "AUC")

Methodclone()

The objects of this class are cloneable with this method.

Usage

Diagnostics_cfg$clone(deep = FALSE)

Arguments

Examples

Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE", "RROC"),   ps = c("SL_risk", "SL_coefs", "AUC"))## ------------------------------------------------## Method `Diagnostics_cfg$new`## ------------------------------------------------Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE", "RROC"),   ps = c("SL_risk", "SL_coefs", "AUC"))## ------------------------------------------------## Method `Diagnostics_cfg$add`## ------------------------------------------------cfg <- Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE", "RROC"),   ps = c("SL_risk", "SL_coefs"))cfg <- cfg$add(ps = "AUC")

Predictor class for the cross-fit predictor of "partial" CATEs

Description

Predictor class for the cross-fit predictor of "partial" CATEs

Predictor class for the cross-fit predictor of "partial" CATEs

Details

The class makes it easier to manage the K predictors for retrieving K-foldcross-validated estimates, as well as to measure how treatment effects changewhen only a single covariate is changed from its "natural" levels (in the sense"natural" used by the direct / indirect effects literature).

Public fields

Methods

Public methods

• FX.Predictor$new()

• FX.Predictor$predict()

• FX.Predictor$clone()

Methodnew()

FX.predictor is a class which simplifies the management of a set of cross-fitprediction models of treatment effects and provides the ability to get the "partial"effects of particular covariates.

Usage

FX.Predictor$new(models, num_splits, num_mc_samples, covariates, model_class)

Arguments

Methodpredict()

Predicts the PCATE surface over a particular covariate, returning a tibble withthe predicted HTE for every Monte-Carlo sample.

Usage

FX.Predictor$predict(data, covariate)

Arguments

Returns

A tibble with columns:

• covariate_value - The value of the covariate of interest

• .hte - An estimated HTE

• .id - The identifier for the original row (which hadcovariate modified tocovariate_value).

Methodclone()

The objects of this class are cloneable with this method.

Usage

FX.Predictor$clone(deep = FALSE)

Arguments

R6 class to represent partitions of the data between training and held-out

Description

R6 class to represent partitions of the data between training and held-out

R6 class to represent partitions of the data between training and held-out

Details

This takes a set of folds calculated elsewhere and representsthese folds in a consistent format.

Public fields

Methods

Public methods

• HTEFold$new()

• HTEFold$clone()

Methodnew()

Creates an R6 object of the data split between training and test set.

Usage

HTEFold$new(data, split_id)

Arguments

Returns

Returns an object of classHTEFold

Methodclone()

The objects of this class are cloneable with this method.

Usage

HTEFold$clone(deep = FALSE)

Arguments

Configuration of Quantities of Interest

Description

HTE_cfg is a configuration class that pulls everything together, indicatingthe full configuration for a given HTE analysis. This includes how to estimatemodels and what Quantities of Interest to calculate based off those underlying models.

Public fields

Methods

Public methods

• HTE_cfg$new()

• HTE_cfg$clone()

Methodnew()

Create a newHTE_cfg object with all necessary information about howto carry out an HTE analysis.

Usage

HTE_cfg$new(  outcome = NULL,  treatment = NULL,  effect = NULL,  qoi = NULL,  verbose = FALSE)

Arguments

Examples

mcate_cfg <- MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))pcate_cfg <- PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))vimp_cfg <- VIMP_cfg$new()diag_cfg <- Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE"),   ps = c("SL_risk", "SL_coefs", "AUC"))qoi_cfg <- QoI_cfg$new(    mcate = mcate_cfg,    pcate = pcate_cfg,    vimp = vimp_cfg,    diag = diag_cfg)ps_cfg <- SLEnsemble_cfg$new(   learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))y_cfg <- SLEnsemble_cfg$new(   learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))fx_cfg <- SLEnsemble_cfg$new(   learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))HTE_cfg$new(outcome = y_cfg, treatment = ps_cfg, effect = fx_cfg, qoi = qoi_cfg)

Methodclone()

The objects of this class are cloneable with this method.

Usage

HTE_cfg$clone(deep = FALSE)

Arguments

Examples

## ------------------------------------------------## Method `HTE_cfg$new`## ------------------------------------------------mcate_cfg <- MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))pcate_cfg <- PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))vimp_cfg <- VIMP_cfg$new()diag_cfg <- Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE"),   ps = c("SL_risk", "SL_coefs", "AUC"))qoi_cfg <- QoI_cfg$new(    mcate = mcate_cfg,    pcate = pcate_cfg,    vimp = vimp_cfg,    diag = diag_cfg)ps_cfg <- SLEnsemble_cfg$new(   learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))y_cfg <- SLEnsemble_cfg$new(   learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))fx_cfg <- SLEnsemble_cfg$new(   learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))HTE_cfg$new(outcome = y_cfg, treatment = ps_cfg, effect = fx_cfg, qoi = qoi_cfg)

Configuration for a Kernel Smoother

Description

KernelSmooth_cfg is a configuration class for non-parametric local-linearregression to construct a smooth representation of the relationship betweentwo variables. This is typically used for displaying a surface of the conditionalaverage treatment effect over a continuous covariate.

Kernel smoothing is handled by thenprobust package.

Super class

tidyhte::Model_cfg ->KernelSmooth_cfg

Public fields

Methods

Public methods

• KernelSmooth_cfg$new()

• KernelSmooth_cfg$clone()

Methodnew()

Create a newKernelSmooth_cfg object with specified number of evaluation points.

Usage

KernelSmooth_cfg$new(neval = 100, eval_min_quantile = 0.05)

Arguments

Returns

A newKernelSmooth_cfg object.

Examples

KernelSmooth_cfg$new(neval = 100)

Methodclone()

The objects of this class are cloneable with this method.

Usage

KernelSmooth_cfg$clone(deep = FALSE)

Arguments

See Also

nprobust::lprobust

Examples

## ------------------------------------------------## Method `KernelSmooth_cfg$new`## ------------------------------------------------KernelSmooth_cfg$new(neval = 100)

Configuration of Known Model

Description

Known_cfg is a configuration class for when a particular model is knowna-priori. The prototypical usage of this class is when heterogeneoustreatment effects are estimated in the context of a randomized controltrial with known propensity scores.

Super class

tidyhte::Model_cfg ->Known_cfg

Public fields

Methods

Public methods

• Known_cfg$new()

• Known_cfg$clone()

Methodnew()

Create a newKnown_cfg object with specified covariate column.

Usage

Known_cfg$new(covariate_name)

Arguments

Returns

A newKnown_cfg object.

Examples

Known_cfg$new("propensity_score")

Methodclone()

The objects of this class are cloneable with this method.

Usage

Known_cfg$clone(deep = FALSE)

Arguments

Examples

## ------------------------------------------------## Method `Known_cfg$new`## ------------------------------------------------Known_cfg$new("propensity_score")

Configuration of Marginal CATEs

Description

MCATE_cfg is a configuration class for estimating marginal responsesurfaces based on heterogeneous treatment effect estimates. "Marginal"in this context implies that all other covariates are marginalized.Thus, if two covariates are highly correlated, it is likely that theirMCATE surfaces will be extremely similar.

Public fields

Methods

Public methods

• MCATE_cfg$new()

• MCATE_cfg$add_moderator()

• MCATE_cfg$clone()

Methodnew()

Create a newMCATE_cfg object with specified model name and hyperparameters.

Usage

MCATE_cfg$new(cfgs, std_errors = TRUE)

Arguments

Returns

A newMCATE_cfg object.

Examples

MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))

Methodadd_moderator()

Add a moderator to theMCATE_cfg object. This entails defining a configurationfor displaying the effect surface for that moderator.

Usage

MCATE_cfg$add_moderator(var_name, cfg)

Arguments

Returns

An updatedMCATE_cfg object.

Examples

cfg <- MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))cfg <- cfg$add_moderator("x2", KernelSmooth_cfg$new(neval = 100))

Methodclone()

The objects of this class are cloneable with this method.

Usage

MCATE_cfg$clone(deep = FALSE)

Arguments

Examples

MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))## ------------------------------------------------## Method `MCATE_cfg$new`## ------------------------------------------------MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))## ------------------------------------------------## Method `MCATE_cfg$add_moderator`## ------------------------------------------------cfg <- MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))cfg <- cfg$add_moderator("x2", KernelSmooth_cfg$new(neval = 100))

Base Class of Model Configurations

Description

Model_cfg is the base class from which all other model configurationsinherit.

Public fields

Methods

Public methods

• Model_cfg$new()

• Model_cfg$clone()

Methodnew()

Create a newModel_cfg object with any necessary parameters.

Usage

Model_cfg$new()

Returns

A newModel_cfg object.

Methodclone()

The objects of this class are cloneable with this method.

Usage

Model_cfg$clone(deep = FALSE)

Arguments

R6 class to represent data to be used in estimating a model

Description

R6 class to represent data to be used in estimating a model

R6 class to represent data to be used in estimating a model

Details

This class provides consistent names and interfaces to data which willbe used in a supervised regression / classification model.

Public fields

Methods

Public methods

• Model_data$new()

• Model_data$SL_cv_control()

• Model_data$clone()

Methodnew()

Creates an R6 object to represent data to be used in a prediction model.

Usage

Model_data$new(data, label_col, ..., .weight_col = NULL)

Arguments

Returns

AModel_data object.

Examples

library("dplyr")df <- dplyr::tibble(    uid = 1:100,    x1 = rnorm(100),    x2 = rnorm(100),    x3 = sample(4, 100, replace = TRUE)) %>% dplyr::mutate(    y = x1 + x2 + x3 + rnorm(100),    x3 = factor(x3))df <- make_splits(df, uid, .num_splits = 5)data <- Model_data$new(df, y, x1, x2, x3)

MethodSL_cv_control()

A helper function to create the cross-validation options to be used by SuperLearner.

Usage

Model_data$SL_cv_control()

Methodclone()

The objects of this class are cloneable with this method.

Usage

Model_data$clone(deep = FALSE)

Arguments

See Also

SuperLearner::SuperLearner.CV.control

Examples

## ------------------------------------------------## Method `Model_data$new`## ------------------------------------------------library("dplyr")df <- dplyr::tibble(    uid = 1:100,    x1 = rnorm(100),    x2 = rnorm(100),    x3 = sample(4, 100, replace = TRUE)) %>% dplyr::mutate(    y = x1 + x2 + x3 + rnorm(100),    x3 = factor(x3))df <- make_splits(df, uid, .num_splits = 5)data <- Model_data$new(df, y, x1, x2, x3)

Configuration of Partial CATEs

Description

PCATE_cfg is a configuration class for estimating marginalresponse surfaces based on heterogeneous treatment effect estimates."Partial" in this context is used similarly to the use in partialdependence plots or in partial regression. In essence, a PCATEattempts to partial out the contribution to the CATE from all othercovariates. Two highly correlated variables may have very differentPCATE surfaces.

Public fields

Methods

Public methods

• PCATE_cfg$new()

• PCATE_cfg$add_moderator()

• PCATE_cfg$clone()

Methodnew()

Create a newPCATE_cfg object with specified model name and hyperparameters.

Usage

PCATE_cfg$new(model_covariates, cfgs, num_mc_samples = 100)

Arguments

Returns

A newPCATE_cfg object.

Examples

PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))

Methodadd_moderator()

Add a moderator to thePCATE_cfg object. This entails adding it to the jointmodel of effects and defines a configuration for displaying the effect surfacefor that moderator.

Usage

PCATE_cfg$add_moderator(var_name, cfg)

Arguments

Returns

An updatedPCATE_cfg object.

Examples

cfg <- PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))cfg <- cfg$add_moderator("x2", KernelSmooth_cfg$new(neval = 100))

Methodclone()

The objects of this class are cloneable with this method.

Usage

PCATE_cfg$clone(deep = FALSE)

Arguments

Examples

PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))## ------------------------------------------------## Method `PCATE_cfg$new`## ------------------------------------------------PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))## ------------------------------------------------## Method `PCATE_cfg$add_moderator`## ------------------------------------------------cfg <- PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))cfg <- cfg$add_moderator("x2", KernelSmooth_cfg$new(neval = 100))

Configuration of Quantities of Interest

Description

QoI_cfg is a configuration class for the Quantities of Interest to begenerated by the HTE analysis.

Public fields

Methods

Public methods

• QoI_cfg$new()

• QoI_cfg$clone()

Methodnew()

Create a newQoI_cfg object with specified Quantities of Interestto estimate.

Usage

QoI_cfg$new(  mcate = NULL,  pcate = NULL,  vimp = NULL,  diag = NULL,  ate = TRUE,  predictions = FALSE)

Arguments

Returns

A newDiagnostics_cfg object.

Examples

mcate_cfg <- MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))pcate_cfg <- PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))vimp_cfg <- VIMP_cfg$new()diag_cfg <- Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE"),   ps = c("SL_risk", "SL_coefs", "AUC"))QoI_cfg$new(    mcate = mcate_cfg,    pcate = pcate_cfg,    vimp = vimp_cfg,    diag = diag_cfg)

Methodclone()

The objects of this class are cloneable with this method.

Usage

QoI_cfg$clone(deep = FALSE)

Arguments

Examples

mcate_cfg <- MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))pcate_cfg <- PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))vimp_cfg <- VIMP_cfg$new()diag_cfg <- Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE"),   ps = c("SL_risk", "SL_coefs", "AUC"))QoI_cfg$new(    mcate = mcate_cfg,    pcate = pcate_cfg,    vimp = vimp_cfg,    diag = diag_cfg)## ------------------------------------------------## Method `QoI_cfg$new`## ------------------------------------------------mcate_cfg <- MCATE_cfg$new(cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)))pcate_cfg <- PCATE_cfg$new(   cfgs = list(x1 = KernelSmooth_cfg$new(neval = 100)),   model_covariates = c("x1", "x2", "x3"),   num_mc_samples = list(x1 = 100))vimp_cfg <- VIMP_cfg$new()diag_cfg <- Diagnostics_cfg$new(   outcome = c("SL_risk", "SL_coefs", "MSE"),   ps = c("SL_risk", "SL_coefs", "AUC"))QoI_cfg$new(    mcate = mcate_cfg,    pcate = pcate_cfg,    vimp = vimp_cfg,    diag = diag_cfg)

Elastic net regression with pairwise interactions

Description

Penalized regression using elastic net. Alpha = 0 corresponds to ridgeregression and alpha = 1 corresponds to Lasso. Included in the modelare pairwise interactions between covariates.

Seevignette("glmnet_beta", package = "glmnet") for a nice tutorial onglmnet.

Usage

SL.glmnet.interaction(  Y,  X,  newX,  family,  obsWeights,  id,  alpha = 1,  nfolds = 10,  nlambda = 100,  useMin = TRUE,  loss = "deviance",  ...)

Arguments

Configuration for a SuperLearner Ensemble

Description

SLEnsemble_cfg is a configuration class for estimation of a modelusing an ensemble of models usingSuperLearner.

Super class

tidyhte::Model_cfg ->SLEnsemble_cfg

Public fields

Methods

Public methods

• SLEnsemble_cfg$new()

• SLEnsemble_cfg$add_sublearner()

• SLEnsemble_cfg$clone()

Methodnew()

Create a newSLEnsemble_cfg object with specified settings.

Usage

SLEnsemble_cfg$new(  cvControl = NULL,  learner_cfgs = NULL,  family = stats::gaussian())

Arguments

Returns

A newSLEnsemble_cfg object.

Examples

SLEnsemble_cfg$new(learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))

Methodadd_sublearner()

Adds a model (or class of models) to the SuperLearner ensemble.If hyperparameter values are specified, this method willadd a learner for every element in the cross-product of providedhyperparameter values.

Usage

SLEnsemble_cfg$add_sublearner(learner_name, hps = NULL)

Arguments

Methodclone()

The objects of this class are cloneable with this method.

Usage

SLEnsemble_cfg$clone(deep = FALSE)

Arguments

Examples

SLEnsemble_cfg$new(learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))## ------------------------------------------------## Method `SLEnsemble_cfg$new`## ------------------------------------------------SLEnsemble_cfg$new(learner_cfgs = list(SLLearner_cfg$new("SL.glm"), SLLearner_cfg$new("SL.gam")))

Configuration of SuperLearner Submodel

Description

SLLearner_cfg is a configuration class for a singlesublearner to be included in SuperLearner. By constructing with a named listof hyperparameters, this configuration allows distinct submodelsfor each unique combination of hyperparameters. To understand what modelsand hyperparameters are available, examine the methods listed inSuperLearner::listWrappers("SL").

Public fields

Methods

Public methods

• SLLearner_cfg$new()

• SLLearner_cfg$clone()

Methodnew()

Create a newSLLearner_cfg object with specified model name and hyperparameters.

Usage

SLLearner_cfg$new(model_name, hp = NULL)

Arguments

Returns

A newSLLearner_cfg object.

Examples

SLLearner_cfg$new("SL.glm")SLLearner_cfg$new("SL.gam", list(deg.gam = c(2, 3)))

Methodclone()

The objects of this class are cloneable with this method.

Usage

SLLearner_cfg$clone(deep = FALSE)

Arguments

Examples

## ------------------------------------------------## Method `SLLearner_cfg$new`## ------------------------------------------------SLLearner_cfg$new("SL.glm")SLLearner_cfg$new("SL.gam", list(deg.gam = c(2, 3)))

Configuration for a Stratification Estimator

Description

Stratified_cfg is a configuration class for stratifying a covariateand calculating statistics within each cell.

Super class

tidyhte::Model_cfg ->Stratified_cfg

Public fields

Methods

Public methods

• Stratified_cfg$new()

• Stratified_cfg$clone()

Methodnew()

Create a newStratified_cfg object with specified number of evaluation points.

Usage

Stratified_cfg$new(covariate)

Arguments

Returns

A newStratified_cfg object.

Examples

Stratified_cfg$new(covariate = "test_covariate")

Methodclone()

The objects of this class are cloneable with this method.

Usage

Stratified_cfg$clone(deep = FALSE)

Arguments

Examples

## ------------------------------------------------## Method `Stratified_cfg$new`## ------------------------------------------------Stratified_cfg$new(covariate = "test_covariate")

Configuration of Variable Importance

Description

VIMP_cfg is a configuration class for estimating a variable importance measureacross all moderators. This provides a meaningful measure of which moderatorsexplain the most of the CATE surface.

Public fields

Methods

Public methods

• VIMP_cfg$new()

• VIMP_cfg$clone()

Methodnew()

Create a newVIMP_cfg object with specified model configuration.

Usage

VIMP_cfg$new(sample_splitting = TRUE, linear_only = FALSE)

Arguments

Returns

A newVIMP_cfg object.

Examples

VIMP_cfg$new()

Methodclone()

The objects of this class are cloneable with this method.

Usage

VIMP_cfg$clone(deep = FALSE)

Arguments

References

• Williamson, B. D., Gilbert, P. B., Carone, M., & Simon, N. (2021).Nonparametric variable importance assessment using machine learning techniques.Biometrics, 77(1), 9-22.

• Williamson, B. D., Gilbert, P. B., Simon, N. R., & Carone, M. (2021).A general framework for inference on algorithm-agnostic variable importance.Journal of the American Statistical Association, 1-14.

Examples

VIMP_cfg$new()## ------------------------------------------------## Method `VIMP_cfg$new`## ------------------------------------------------VIMP_cfg$new()

Add an additional diagnostic to the effect model

Description

This adds a diagnostic to the effect model.

Usage

add_effect_diagnostic(hte_cfg, diag)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_effect_diagnostic("RROC") -> hte_cfg

Add an additional model to the joint effect ensemble

Description

This adds a learner to the ensemble used for estimating a modelof the conditional expectation of the pseudo-outcome.

Usage

add_effect_model(hte_cfg, model_name, ...)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_effect_model("SL.glm.interaction") -> hte_cfg

Uses a known propensity score

Description

This replaces the propensity score model with a known valueof the propensity score.

Usage

add_known_propensity_score(hte_cfg, covariate_name)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_known_propensity_score("ps") -> hte_cfg

Adds moderators to the configuration

Description

This adds a definition about how to display a moderators tothe MCATE config. A moderator is any variable that you want to view informationabout CATEs with respect to.

Usage

add_moderator(hte_cfg, model_type, ..., .model_arguments = NULL)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_moderator("Stratified", x2, x3) %>%   add_moderator("KernelSmooth", x1, x4, x5) -> hte_cfg

Add an additional diagnostic to the outcome model

Description

This adds a diagnostic to the outcome model.

Usage

add_outcome_diagnostic(hte_cfg, diag)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_outcome_diagnostic("RROC") -> hte_cfg

Add an additional model to the outcome ensemble

Description

This adds a learner to the ensemble used for estimating a modelof the conditional expectation of the outcome.

Usage

add_outcome_model(hte_cfg, model_name, ...)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_outcome_model("SL.glm.interaction") -> hte_cfg

Add an additional diagnostic to the propensity score

Description

This adds a diagnostic to the propensity score.

Usage

add_propensity_diagnostic(hte_cfg, diag)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_propensity_diagnostic(c("AUC", "MSE")) -> hte_cfg

Add an additional model to the propensity score ensemble

Description

This adds a learner to the ensemble used for estimating propensityscores.

Usage

add_propensity_score_model(hte_cfg, model_name, ...)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   add_propensity_score_model("SL.glmnet", alpha = c(0, 0.5, 1)) -> hte_cfg

Adds variable importance information

Description

This adds a variable importance quantity of interest to the outputs.

Usage

add_vimp(hte_cfg, sample_splitting = TRUE, linear_only = FALSE)

Arguments

Value

UpdatedHTE_cfg object

References

• Williamson, B. D., Gilbert, P. B., Carone, M., & Simon, N. (2021).Nonparametric variable importance assessment using machine learning techniques.Biometrics, 77(1), 9-22.

• Williamson, B. D., Gilbert, P. B., Simon, N. R., & Carone, M. (2021).A general framework for inference on algorithm-agnostic variable importance.Journal of the American Statistical Association, 1-14.

Examples

library("dplyr")basic_config() %>%   add_vimp(sample_splitting = FALSE) -> hte_cfg

Attach anHTE_cfg to a dataframe

Description

This adds a configuration attribute to a dataframe for HTE estimation.This configuration details the full analysis of HTE that should be performed.

Usage

attach_config(data, .HTE_cfg)

Arguments

Details

For information about how to set up anHTE_cfg object, see the Recipe APIdocumentationbasic_config().

To see an example analysis, readvignette("experimental_analysis") in the contextof an experiment,vignette("experimental_analysis") for an observational study, orvignette("methodological_details") for a deeper dive under the hood.

See Also

basic_config(),make_splits(),produce_plugin_estimates(),construct_pseudo_outcomes(),estimate_QoI()

Examples

library("dplyr")if(require("palmerpenguins")) {data(package = 'palmerpenguins')penguins$unitid = seq_len(nrow(penguins))penguins$propensity = rep(0.5, nrow(penguins))penguins$treatment = rbinom(nrow(penguins), 1, penguins$propensity)cfg <- basic_config() %>% add_known_propensity_score("propensity") %>%add_outcome_model("SL.glm.interaction") %>%remove_vimp()attach_config(penguins, cfg) %>%make_splits(unitid, .num_splits = 4) %>%produce_plugin_estimates(outcome = body_mass_g, treatment = treatment, species, sex) %>%construct_pseudo_outcomes(body_mass_g, treatment) %>%estimate_QoI(species, sex)}

Create a basic config for HTE estimation

Description

This provides a basic recipe for HTE estimation that canbe extended by providing additional information about modelsto be estimated and what quantities of interest should bereturned based on those models. This basic model includesonly linear models for nuisance function estimation, andbasic diagnostics.

Usage

basic_config()

Details

Additional models, diagnostics and quantities of interest shouldbe added using their respective helper functions provided as partof the Recipe API.

To see an example analysis, readvignette("experimental_analysis") in the contextof an experiment,vignette("experimental_analysis") for an observational study, orvignette("methodological_details") for a deeper dive under the hood.

Value

HTE_cfg object

See Also

add_propensity_score_model(),add_known_propensity_score(),add_propensity_diagnostic(),add_outcome_model(),add_outcome_diagnostic(),add_effect_model(),add_effect_diagnostic(),add_moderator(),add_vimp()

Examples

library("dplyr")basic_config() %>%   add_known_propensity_score("ps") %>%   add_outcome_model("SL.glm.interaction") %>%   add_outcome_model("SL.glmnet", alpha = c(0.05, 0.15, 0.2, 0.25, 0.5, 0.75)) %>%   add_outcome_model("SL.glmnet.interaction", alpha = c(0.05, 0.15, 0.2, 0.25, 0.5, 0.75)) %>%   add_outcome_diagnostic("RROC") %>%   add_effect_model("SL.glm.interaction") %>%   add_effect_model("SL.glmnet", alpha = c(0.05, 0.15, 0.2, 0.25, 0.5, 0.75)) %>%   add_effect_model("SL.glmnet.interaction", alpha = c(0.05, 0.15, 0.2, 0.25, 0.5, 0.75)) %>%   add_effect_diagnostic("RROC") %>%   add_moderator("Stratified", x2, x3) %>%   add_moderator("KernelSmooth", x1, x4, x5) %>%   add_vimp(sample_splitting = FALSE) -> hte_cfg

Calculates a SATE and a PATE using AIPW

Description

This function takes fully prepared data (with all auxilliary columns from thenecessary models) and estimates average treatment effects using AIPW.

Usage

calculate_ate(data)

Arguments

References

• Kennedy, E. H. (2020). Towards optimal doubly robust estimation of heterogeneouscausal effects.arXiv preprint arXiv:2004.14497.

• Tsiatis, A. A., Davidian, M., Zhang, M., & Lu, X. (2008). Covariate adjustmentfor two‐sample treatment comparisons in randomized clinical trials: a principledyet flexible approach.Statistics in medicine, 27(23), 4658-4677.

See Also

basic_config(),attach_config(),make_splits(),produce_plugin_estimates(),construct_pseudo_outcomes(),estimate_QoI()

Calculate diagnostics

Description

This function calculates the diagnostics requested by theDiagnostics_cfg object.

Usage

calculate_diagnostics(data, treatment, outcome, .diag.cfg)

Arguments

Value

Returns a tibble with columns:

• estimand - Character indicating the diagnostic that was calculated

• level - Indicates the scope of this diagnostic (e.g. does it applyonly to the model of the outcome under treatment).

• term - Indicates a more granular descriptor of what the value is for,such as the specific model within the SuperLearner ensemble.

• estimate - Point estimate of the diagnostic.

• std_error - Standard error of the diagnostic.

See Also

Diagnostics_cfg

Calculate Linear Variable Importance of HTEs

Description

calculate_linear_vimp estimates the linear hypothesis test of removing a particular moderatorfrom a linear model containing all moderators. Unlikecalculate_vimp, this will only beunbiased and have correct asymptotic coverage rates if the true model is linear. This linearapproach is also substantially faster, so may be useful when prototyping an analysis.

Usage

calculate_linear_vimp(  full_data,  weight_col,  pseudo_outcome,  ...,  .VIMP_cfg,  .Model_cfg)

Arguments

References

• Williamson, B. D., Gilbert, P. B., Carone, M., & Simon, N. (2021).Nonparametric variable importance assessment using machine learning techniques.Biometrics, 77(1), 9-22.

• Williamson, B. D., Gilbert, P. B., Simon, N. R., & Carone, M. (2021).A general framework for inference on algorithm-agnostic variable importance.Journal of the American Statistical Association, 1-14.

See Also

calculate_vimp()

Calculate "partial" CATE estimates

Description

Usage

calculate_pcate_quantities(  full_data,  .weights,  .outcome,  fx_model,  ...,  .MCATE_cfg)

Regression ROC Curve calculation

Description

This function calculates the RegressionROC Curve ofof Hernández-Orallodoi:10.1016/j.patcog.2013.06.014.It provides estimates for the positive and negativeerrors when predictions are shifted by a varietyof constants (which range across the domain of observedresiduals). Curves closer to the axes are, in general, to bepreferred. In general, this curve provides a simple way tovisualize the error properties of a regression model.

Usage

calculate_rroc(label, prediction, nbins = 100)

Arguments

Details

The dot shows the errors when no shift is applied, correspondingto the base model predictions.

Value

A tibble withnbins rows.

References

Hernández-Orallo, J. (2013). ROC curves for regression.Pattern Recognition, 46(12), 3395-3411.

Calculate Variable Importance of HTEs

Description

calculate_vimp estimates the reduction in (population) $R^2$ fromremoving a particular moderator from a model containing all moderators.

Usage

calculate_vimp(  full_data,  weight_col,  pseudo_outcome,  ...,  .VIMP_cfg,  .Model_cfg)

Arguments

References

• Williamson, B. D., Gilbert, P. B., Carone, M., & Simon, N. (2021).Nonparametric variable importance assessment using machine learning techniques.Biometrics, 77(1), 9-22.

• Williamson, B. D., Gilbert, P. B., Simon, N. R., & Carone, M. (2021).A general framework for inference on algorithm-agnostic variable importance.Journal of the American Statistical Association, 1-14.

See Also

calculate_linear_vimp()

Checks that a dataframe has an attached configuration for HTEs

Description

This helper function ensures that the provided dataframe hasthe necessary auxilliary configuration information for HTEestimation.

Usage

check_data_has_hte_cfg(data)

Arguments

Value

Returns NULL. Errors if a problem is discovered.

Checks that an appropriate identifier has been provided

Description

This helper function makes a few simple checks to identify obviousissues with the way provided column of unit identifiers.

Usage

check_identifier(data, id_col)

Arguments

Value

Returns NULL. Errors if a problem is discovered.

Checks that nuisance models have been estimated and exist in the supplied dataset.

Description

This helper function makes a few simple checks to identify obviousissues with the way that nuisance functions are created and prepared.

Usage

check_nuisance_models(data)

Arguments

Value

Returns NULL. Errors if a problem is discovered.

Checks that splits have been properly created.

Description

This helper function makes a few simple checks to identify obviousissues with the way that splits have been made in the supplied data.

Usage

check_splits(data)

Arguments

Value

Returns NULL. Errors if a problem is discovered.

Checks that an appropriate weighting variable has been provided

Description

This helper function makes a few simple checks to identify obviousissues with the weights provided.

Usage

check_weights(data, weight_col)

Arguments

Value

Returns NULL. Errors if a problem is discovered.

Construct Pseudo-outcomes

Description

construct_pseudo_outcomes takes a dataset which has been preparedwith plugin estimators of nuisance parameters and transforms these intoa "pseudo-outcome": an unbiased estimator of the conditional averagetreatment effect under exogeneity.

Usage

construct_pseudo_outcomes(data, outcome, treatment, type = "dr")

Arguments

Details

Taking averages of these pseudo-outcomes (or fitting a model to them)will approximate averages (or models) of the underlying treatment effect.

See Also

attach_config(),make_splits(),produce_plugin_estimates(),estimate_QoI()

Estimate Quantities of Interest

Description

estimate_QoI takes a dataframe already prepared with split IDs,plugin estimates and pseudo-outcomes and calculates the requestedquantities of interest (QoIs).

Usage

estimate_QoI(data, ...)

Arguments

Details

To see an example analysis, readvignette("experimental_analysis") in the contextof an experiment,vignette("experimental_analysis") for an observational study, orvignette("methodological_details") for a deeper dive under the hood.

See Also

attach_config(),make_splits(),produce_plugin_estimates(),construct_pseudo_outcomes(),

Examples

library("dplyr")if(require("palmerpenguins")) {data(package = 'palmerpenguins')penguins$unitid = seq_len(nrow(penguins))penguins$propensity = rep(0.5, nrow(penguins))penguins$treatment = rbinom(nrow(penguins), 1, penguins$propensity)cfg <- basic_config() %>% add_known_propensity_score("propensity") %>%add_outcome_model("SL.glm.interaction") %>%remove_vimp()attach_config(penguins, cfg) %>%make_splits(unitid, .num_splits = 4) %>%produce_plugin_estimates(outcome = body_mass_g, treatment = treatment, species, sex) %>%construct_pseudo_outcomes(body_mass_g, treatment) %>%estimate_QoI(species, sex)}

Function to calculate diagnostics based on model outputs

Description

This function defines the calculations of common model diagnosticswhich are available.

Usage

estimate_diagnostic(data, label, prediction, diag_name, params)

Arguments

Examples

df <- dplyr::tibble(y = rbinom(100, 1, 0.5), p = rep(0.5, 100), w = rexp(100), u = 1:100)attr(df, "weights") <- "w"attr(df, "identifier") <- "u"estimate_diagnostic(df, "y", "p", "AUC")

Fits a treatment effect model using the appropriate settings

Description

This function prepares data, fits the appropriate model and returns theresulting estimates in a standardized format.

Usage

fit_effect(full_data, weight_col, fx_col, ..., .Model_cfg)

Arguments

Value

A list with one element,fx. This element contains aPredictor object ofthe appropriate subclass corresponding to theModel_cfg fit to the data.

Fit a predictor for treatment effects

Description

This function predicts treatment effects in a second stage model.

Usage

fit_fx_predictor(full_data, weights, psi_col, ..., .pcate.cfg, .Model_cfg)

Arguments

Value

A list with two items:

• model - TheFX.Predictor model object used internally for PCATE estimation.

• data - The data augmented with column.pseudo_outcome_hat for the cross-fit predictionsof the HTE for each unit.

See Also

Model_cfg,PCATE_cfg

Fits a plugin model using the appropriate settings

Description

This function prepares data, fits the appropriate models and returns theresulting estimates in a standardized format.

Usage

fit_plugin(full_data, weight_col, outcome_col, ..., .Model_cfg)

Arguments

Value

A newPredictor object of the appropriate subclass corresponding to theModel_cfg fit to the data.

Fits a propensity score model using the appropriate settings

Description

This function prepares data, fits the appropriate model and returns theresulting estimates in a standardized format.

Usage

fit_plugin_A(full_data, weight_col, a_col, ..., .Model_cfg)

Arguments

Value

A list with one element,ps. This element contains aPredictor object ofthe appropriate subclass corresponding to theModel_cfg fit to the data.

Fits a T-learner using the appropriate settings

Description

This function prepares data, fits the appropriate model and returns theresulting estimates in a standardized format.

Usage

fit_plugin_Y(full_data, weight_col, y_col, a_col, ..., .Model_cfg)

Arguments

Value

A list with two elements,mu1 andmu0 corresponding to the models fit tothe treatment and control potential outcomes, respectively. Each is a newPredictorobject of the appropriate subclass corresponding to the theModel_cfg fit to the data.

Removes rows which have missing data on any of the supplied columns.

Description

This function removes rows with missingness based on the columns provided.If rows are dropped, a message is displayed to the user to inform them of thisfact.

Usage

listwise_deletion(data, ...)

Arguments

Value

The original data with all observations which are fully observed.

Define splits for cross-fitting

Description

This takes a dataset, a column with a unique identifier and anarbitrary number of covariates on which to stratify the splits.It returns the original dataset with an additional column.split_idcorresponding to an identifier for the split.

Usage

make_splits(data, identifier, ..., .num_splits)

Arguments

Details

To see an example analysis, readvignette("experimental_analysis") in the contextof an experiment,vignette("experimental_analysis") for an observational study, orvignette("methodological_details") for a deeper dive under the hood.

Value

original dataframe with additional.split_id column

See Also

attach_config(),produce_plugin_estimates(),construct_pseudo_outcomes(),estimate_QoI()

Examples

library("dplyr")if(require("palmerpenguins")) {data(package = 'palmerpenguins')penguins$unitid = seq_len(nrow(penguins))penguins$propensity = rep(0.5, nrow(penguins))penguins$treatment = rbinom(nrow(penguins), 1, penguins$propensity)cfg <- basic_config() %>% add_known_propensity_score("propensity") %>%add_outcome_model("SL.glm.interaction") %>%remove_vimp()attach_config(penguins, cfg) %>%make_splits(unitid, .num_splits = 4) %>%produce_plugin_estimates(outcome = body_mass_g, treatment = treatment, species, sex) %>%construct_pseudo_outcomes(body_mass_g, treatment) %>%estimate_QoI(species, sex)}

Prediction for an SL.glmnet object

Description

Prediction for the glmnet wrapper.

Usage

## S3 method for class 'SL.glmnet.interaction'predict(  object,  newdata,  remove_extra_cols = TRUE,  add_missing_cols = TRUE,  ...)

Arguments

See Also

SuperLearner::SL.glmnet

Estimate models of nuisance functions

Description

This takes a dataset with an identified outcome and treatment column alongwith any number of covariates and appends three columns to the dataset correspondingto an estimate of the conditional expectation of treatment (.pi_hat), along with theconditional expectation of the control and treatment potential outcome surfaces(.mu0_hat and.mu1_hat respectively).

Usage

produce_plugin_estimates(data, outcome, treatment, ..., .weights = NULL)

Arguments

Details

To see an example analysis, readvignette("experimental_analysis") in the contextof an experiment,vignette("experimental_analysis") for an observational study, orvignette("methodological_details") for a deeper dive under the hood.

See Also

attach_config(),make_splits(),construct_pseudo_outcomes(),estimate_QoI()

Examples

library("dplyr")if(require("palmerpenguins")) {data(package = 'palmerpenguins')penguins$unitid = seq_len(nrow(penguins))penguins$propensity = rep(0.5, nrow(penguins))penguins$treatment = rbinom(nrow(penguins), 1, penguins$propensity)cfg <- basic_config() %>% add_known_propensity_score("propensity") %>%add_outcome_model("SL.glm.interaction") %>%remove_vimp()attach_config(penguins, cfg) %>%make_splits(unitid, .num_splits = 4) %>%produce_plugin_estimates(outcome = body_mass_g, treatment = treatment, species, sex) %>%construct_pseudo_outcomes(body_mass_g, treatment) %>%estimate_QoI(species, sex)}

Removes variable importance information

Description

This removes the variable importance quantity of interestfrom anHTE_cfg.

Usage

remove_vimp(hte_cfg)

Arguments

Value

UpdatedHTE_cfg object

Examples

library("dplyr")basic_config() %>%   remove_vimp() -> hte_cfg

Partition the data into folds

Description

This takes a dataset and a split ID and generates two subsets of thedata corresponding to a training set and a holdout.

Usage

split_data(data, split_id)

Arguments

Value

Returns an R6 objectHTEFold with three public fields:

• train - The split to be used for training the plugin estimates

• holdout - The split not used for training

• in_holdout - A logical vector indicating for each unit whether they lie in the holdout.