Package 'DoubleML'

Description

Initalization of DoubleMLData from data.frame.

Usage

double_ml_data_from_data_frame(
  df,
  x_cols = NULL,
  y_col = NULL,
  d_cols = NULL,
  z_cols = NULL,
  cluster_cols = NULL,
  use_other_treat_as_covariate = TRUE
)

Arguments

Value

Creates a new instance of class DoubleMLData.

Examples

df = make_plr_CCDDHNR2018(return_type = "data.frame")
x_names = names(df)[grepl("X", names(df))]
obj_dml_data = double_ml_data_from_data_frame(
  df = df, x_cols = x_names,
  y_col = "y", d_cols = "d")
# Input: Data frame, Output: DoubleMLData object

Description

Initalization of DoubleMLData from matrix() objects.

Usage

double_ml_data_from_matrix(
  X = NULL,
  y,
  d,
  z = NULL,
  cluster_vars = NULL,
  data_class = "DoubleMLData",
  use_other_treat_as_covariate = TRUE
)

Arguments

Value

Creates a new instance of class DoubleMLData.

Examples

matrix_list = make_plr_CCDDHNR2018(return_type = "matrix")
obj_dml_data = double_ml_data_from_matrix(
  X = matrix_list$X,
  y = matrix_list$y,
  d = matrix_list$d)

Description

Abstract base class that can't be initialized.

Format

R6::R6Class object.

Active bindings

Methods

Public methods

• DoubleML$new()

• DoubleML$print()

• DoubleML$fit()

• DoubleML$bootstrap()

• DoubleML$split_samples()

• DoubleML$set_sample_splitting()

• DoubleML$tune()

• DoubleML$summary()

• DoubleML$confint()

• DoubleML$learner_names()

• DoubleML$params_names()

• DoubleML$set_ml_nuisance_params()

• DoubleML$p_adjust()

• DoubleML$get_params()

• DoubleML$clone()

Method new()

DoubleML is an abstract class that can't be initialized.

Usage

DoubleML$new()

Method print()

Print DoubleML objects.

Usage

DoubleML$print()

Method fit()

Estimate DoubleML models.

Usage

DoubleML$fit(store_predictions = FALSE, store_models = FALSE)

Arguments

Returns

self

Method bootstrap()

Multiplier bootstrap for DoubleML models.

Usage

DoubleML$bootstrap(method = "normal", n_rep_boot = 500)

Arguments

Returns

self

Method split_samples()

Draw sample splitting for DoubleML models.

The samples are drawn according to the attributes n_folds, n_rep and apply_cross_fitting.

Usage

DoubleML$split_samples()

Returns

self

Method set_sample_splitting()

Set the sample splitting for DoubleML models.

The attributes n_folds and n_rep are derived from the provided partition.

Usage

DoubleML$set_sample_splitting(smpls)

Arguments

Returns

self

Examples

library(DoubleML)
library(mlr3)
set.seed(2)
obj_dml_data = make_plr_CCDDHNR2018(n_obs=10)
dml_plr_obj = DoubleMLPLR$new(obj_dml_data,
                              lrn("regr.rpart"), lrn("regr.rpart"))

# simple sample splitting with two folds and without cross-fitting
smpls = list(list(train_ids = list(c(1, 2, 3, 4, 5)),
                  test_ids = list(c(6, 7, 8, 9, 10))))
dml_plr_obj$set_sample_splitting(smpls)

# sample splitting with two folds and cross-fitting but no repeated cross-fitting
smpls = list(list(train_ids = list(c(1, 2, 3, 4, 5), c(6, 7, 8, 9, 10)),
                  test_ids = list(c(6, 7, 8, 9, 10), c(1, 2, 3, 4, 5))))
dml_plr_obj$set_sample_splitting(smpls)

# sample splitting with two folds and repeated cross-fitting with n_rep = 2
smpls = list(list(train_ids = list(c(1, 2, 3, 4, 5), c(6, 7, 8, 9, 10)),
                  test_ids = list(c(6, 7, 8, 9, 10), c(1, 2, 3, 4, 5))),
             list(train_ids = list(c(1, 3, 5, 7, 9), c(2, 4, 6, 8, 10)),
                  test_ids = list(c(2, 4, 6, 8, 10), c(1, 3, 5, 7, 9))))
dml_plr_obj$set_sample_splitting(smpls)

Method tune()

Hyperparameter-tuning for DoubleML models.

The hyperparameter-tuning is performed using the tuning methods provided in the mlr3tuning package. For more information on tuning in mlr3, we refer to the section on parameter tuning in the mlr3 book.

Usage

DoubleML$tune(
  param_set,
  tune_settings = list(n_folds_tune = 5, rsmp_tune = mlr3::rsmp("cv", folds = 5), measure
    = NULL, terminator = mlr3tuning::trm("evals", n_evals = 20), algorithm =
    mlr3tuning::tnr("grid_search"), resolution = 5),
  tune_on_folds = FALSE
)

Arguments

Returns

self

Method summary()

Summary for DoubleML models after calling fit().

Usage

DoubleML$summary(digits = max(3L, getOption("digits") - 3L))

Arguments

Method confint()

Confidence intervals for DoubleML models.

Usage

DoubleML$confint(parm, joint = FALSE, level = 0.95)

Arguments

Returns

A matrix() with the confidence interval(s).

Method learner_names()

Returns the names of the learners.

Usage

DoubleML$learner_names()

Returns

character() with names of learners.

Method params_names()

Returns the names of the nuisance models with hyperparameters.

Usage

DoubleML$params_names()

Returns

character() with names of nuisance models with hyperparameters.

Method set_ml_nuisance_params()

Set hyperparameters for the nuisance models of DoubleML models.

Note that in the current implementation, either all parameters have to be set globally or all parameters have to be provided fold-specific.

Usage

DoubleML$set_ml_nuisance_params(
  learner = NULL,
  treat_var = NULL,
  params,
  set_fold_specific = FALSE
)

Arguments

Returns

self

Method p_adjust()

Multiple testing adjustment for DoubleML models.

Usage

DoubleML$p_adjust(method = "romano-wolf", return_matrix = TRUE)

Arguments

Returns

numeric() with adjusted p-values. If return_matrix = TRUE, a matrix() with adjusted p_values.

Method get_params()

Get hyperparameters for the nuisance model of DoubleML models.

Usage

DoubleML$get_params(learner)

Arguments

Returns

named list()with paramers for the nuisance model/learner.

Method clone()

The objects of this class are cloneable with this method.

Usage

DoubleML$clone(deep = FALSE)

Arguments

See Also

Other DoubleML: DoubleMLIIVM, DoubleMLIRM, DoubleMLPLIV, DoubleMLPLR

Examples

## ------------------------------------------------
## Method `DoubleML$set_sample_splitting`
## ------------------------------------------------

library(DoubleML)
library(mlr3)
set.seed(2)
obj_dml_data = make_plr_CCDDHNR2018(n_obs=10)
dml_plr_obj = DoubleMLPLR$new(obj_dml_data,
                              lrn("regr.rpart"), lrn("regr.rpart"))

# simple sample splitting with two folds and without cross-fitting
smpls = list(list(train_ids = list(c(1, 2, 3, 4, 5)),
                  test_ids = list(c(6, 7, 8, 9, 10))))
dml_plr_obj$set_sample_splitting(smpls)

# sample splitting with two folds and cross-fitting but no repeated cross-fitting
smpls = list(list(train_ids = list(c(1, 2, 3, 4, 5), c(6, 7, 8, 9, 10)),
                  test_ids = list(c(6, 7, 8, 9, 10), c(1, 2, 3, 4, 5))))
dml_plr_obj$set_sample_splitting(smpls)

# sample splitting with two folds and repeated cross-fitting with n_rep = 2
smpls = list(list(train_ids = list(c(1, 2, 3, 4, 5), c(6, 7, 8, 9, 10)),
                  test_ids = list(c(6, 7, 8, 9, 10), c(1, 2, 3, 4, 5))),
             list(train_ids = list(c(1, 3, 5, 7, 9), c(2, 4, 6, 8, 10)),
                  test_ids = list(c(2, 4, 6, 8, 10), c(1, 3, 5, 7, 9))))
dml_plr_obj$set_sample_splitting(smpls)

Description

Double machine learning data-backend for data with cluster variables.

DoubleMLClusterData objects can be initialized from a data.table. Alternatively DoubleML provides functions to initialize from a collection of matrix objects or a data.frame. The following functions can be used to create a new instance of DoubleMLClusterData.

• DoubleMLClusterData$new() for initialization from a data.table.

• double_ml_data_from_matrix() for initialization from matrix objects,

• double_ml_data_from_data_frame() for initialization from a data.frame.

Super class

DoubleML::DoubleMLData -> DoubleMLClusterData

Active bindings

Methods

Public methods

• DoubleMLClusterData$new()

• DoubleMLClusterData$print()

• DoubleMLClusterData$set_data_model()

• DoubleMLClusterData$clone()

Method new()

Creates a new instance of this R6 class.

Usage

DoubleMLClusterData$new(
  data = NULL,
  x_cols = NULL,
  y_col = NULL,
  d_cols = NULL,
  cluster_cols = NULL,
  z_cols = NULL,
  use_other_treat_as_covariate = TRUE
)

Arguments

Method print()

Print DoubleMLClusterData objects.

Usage

DoubleMLClusterData$print()

Method set_data_model()

Setter function for data_model. The function implements the causal model as specified by the user via y_col, d_cols, x_cols, z_cols and cluster_cols and assigns the role for the treatment variables in the multiple-treatment case.

Usage

DoubleMLClusterData$set_data_model(treatment_var)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

DoubleMLClusterData$clone(deep = FALSE)

Arguments

Examples

library(DoubleML)
dt = make_pliv_multiway_cluster_CKMS2021(return_type = "data.table")
obj_dml_data = DoubleMLClusterData$new(dt,
  y_col = "Y",
  d_cols = "D",
  z_cols = "Z",
  cluster_cols = c("cluster_var_i", "cluster_var_j"))

Description

Double machine learning data-backend.

DoubleMLData objects can be initialized from a data.table. Alternatively DoubleML provides functions to initialize from a collection of matrix objects or a data.frame. The following functions can be used to create a new instance of DoubleMLData.

• DoubleMLData$new() for initialization from a data.table.

• double_ml_data_from_matrix() for initialization from matrix objects,

• double_ml_data_from_data_frame() for initialization from a data.frame.

Active bindings

Methods

Public methods

• DoubleMLData$new()

• DoubleMLData$print()

• DoubleMLData$set_data_model()

• DoubleMLData$clone()

Method new()

Creates a new instance of this R6 class.

Usage

DoubleMLData$new(
  data = NULL,
  x_cols = NULL,
  y_col = NULL,
  d_cols = NULL,
  z_cols = NULL,
  use_other_treat_as_covariate = TRUE
)

Arguments

Method print()

Print DoubleMLData objects.

Usage

DoubleMLData$print()

Method set_data_model()

Setter function for data_model. The function implements the causal model as specified by the user via y_col, d_cols, x_cols and z_cols and assigns the role for the treatment variables in the multiple-treatment case.

Usage

DoubleMLData$set_data_model(treatment_var)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

DoubleMLData$clone(deep = FALSE)

Arguments

Examples

library(DoubleML)
df = make_plr_CCDDHNR2018(return_type = "data.table")
obj_dml_data = DoubleMLData$new(df,
  y_col = "y",
  d_cols = "d")

Description

Double machine learning for interactive IV regression models.

Format

R6::R6Class object inheriting from DoubleML.

Details

Interactive IV regression (IIVM) models take the form

$Y = \ell_0(D,X) + \zeta$,

$Z = m_0(X) + V$,

with $E[\zeta|X,Z]=0$ and $E[V|X] = 0$. $Y$ is the outcome variable, $D \in \{0,1\}$ is the binary treatment variable and $Z \in \{0,1\}$ is a binary instrumental variable. Consider the functions $g_0$, $r_0$ and $m_0$, where $g_0$ maps the support of $(Z,X)$ to $R$ and $r_0$ and $m_0$, respectively, map the support of $(Z,X)$ and $X$ to $(\epsilon, 1-\epsilon)$ for some $\epsilon \in (1, 1/2)$, such that

$Y = g_0(Z,X) + \nu,$

$D = r_0(Z,X) + U,$

$Z = m_0(X) + V,$

with $E[\nu|Z,X]=0$, $E[U|Z,X]=0$ and $E[V|X]=0$. The target parameter of interest in this model is the local average treatment effect (LATE),

$\theta_0 = \frac{E[g_0(1,X)] - E[g_0(0,X)]}{E[r_0(1,X)] - E[r_0(0,X)]}.$

Super class

DoubleML::DoubleML -> DoubleMLIIVM

Active bindings

Methods

Public methods

• DoubleMLIIVM$new()

• DoubleMLIIVM$clone()

Method new()

Creates a new instance of this R6 class.

Usage

DoubleMLIIVM$new(
  data,
  ml_g,
  ml_m,
  ml_r,
  n_folds = 5,
  n_rep = 1,
  score = "LATE",
  subgroups = list(always_takers = TRUE, never_takers = TRUE),
  dml_procedure = "dml2",
  trimming_rule = "truncate",
  trimming_threshold = 1e-12,
  draw_sample_splitting = TRUE,
  apply_cross_fitting = TRUE
)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

DoubleMLIIVM$clone(deep = FALSE)

Arguments

See Also

Other DoubleML: DoubleML, DoubleMLIRM, DoubleMLPLIV, DoubleMLPLR

Examples

library(DoubleML)
library(mlr3)
library(mlr3learners)
library(data.table)
set.seed(2)
ml_g = lrn("regr.ranger",
  num.trees = 100, mtry = 20,
  min.node.size = 2, max.depth = 5)
ml_m = lrn("classif.ranger",
  num.trees = 100, mtry = 20,
  min.node.size = 2, max.depth = 5)
ml_r = ml_m$clone()
obj_dml_data = make_iivm_data(
  theta = 0.5, n_obs = 1000,
  alpha_x = 1, dim_x = 20)
dml_iivm_obj = DoubleMLIIVM$new(obj_dml_data, ml_g, ml_m, ml_r)
dml_iivm_obj$fit()
dml_iivm_obj$summary()


## Not run: 
library(DoubleML)
library(mlr3)
library(mlr3learners)
library(mlr3tuning)
library(data.table)
set.seed(2)
ml_g = lrn("regr.rpart")
ml_m = lrn("classif.rpart")
ml_r = ml_m$clone()
obj_dml_data = make_iivm_data(
  theta = 0.5, n_obs = 1000,
  alpha_x = 1, dim_x = 20)
dml_iivm_obj = DoubleMLIIVM$new(obj_dml_data, ml_g, ml_m, ml_r)
param_grid = list(
  "ml_g" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)),
  "ml_m" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)),
  "ml_r" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)))
# minimum requirements for tune_settings
tune_settings = list(
  terminator = mlr3tuning::trm("evals", n_evals = 5),
  algorithm = mlr3tuning::tnr("grid_search", resolution = 5))
dml_iivm_obj$tune(param_set = param_grid, tune_settings = tune_settings)
dml_iivm_obj$fit()
dml_iivm_obj$summary()

## End(Not run)

Description

Double machine learning for interactive regression models.

Format

R6::R6Class object inheriting from DoubleML.

Details

Interactive regression (IRM) models take the form

$Y = g_0(D,X) + U$,

$D = m_0(X) + V$,

with $E[U|X,D]=0$ and $E[V|X] = 0$. $Y$ is the outcome variable and $D \in \{0,1\}$ is the binary treatment variable. We consider estimation of the average treamtent effects when treatment effects are fully heterogeneous. Target parameters of interest in this model are the average treatment effect (ATE),

$\theta_0 = E[g_0(1,X) - g_0(0,X)]$

and the average treament effect on the treated (ATTE),

$\theta_0 = E[g_0(1,X) - g_0(0,X)|D=1]$.

Super class

DoubleML::DoubleML -> DoubleMLIRM

Active bindings

Methods

Public methods

• DoubleMLIRM$new()

• DoubleMLIRM$clone()

Method new()

Creates a new instance of this R6 class.

Usage

DoubleMLIRM$new(
  data,
  ml_g,
  ml_m,
  n_folds = 5,
  n_rep = 1,
  score = "ATE",
  trimming_rule = "truncate",
  trimming_threshold = 1e-12,
  dml_procedure = "dml2",
  draw_sample_splitting = TRUE,
  apply_cross_fitting = TRUE
)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

DoubleMLIRM$clone(deep = FALSE)

Arguments

See Also

Other DoubleML: DoubleML, DoubleMLIIVM, DoubleMLPLIV, DoubleMLPLR

Examples

library(DoubleML)
library(mlr3)
library(mlr3learners)
library(data.table)
set.seed(2)
ml_g = lrn("regr.ranger",
  num.trees = 100, mtry = 20,
  min.node.size = 2, max.depth = 5)
ml_m = lrn("classif.ranger",
  num.trees = 100, mtry = 20,
  min.node.size = 2, max.depth = 5)
obj_dml_data = make_irm_data(theta = 0.5)
dml_irm_obj = DoubleMLIRM$new(obj_dml_data, ml_g, ml_m)
dml_irm_obj$fit()
dml_irm_obj$summary()

## Not run: 
library(DoubleML)
library(mlr3)
library(mlr3learners)
library(mlr3uning)
library(data.table)
set.seed(2)
ml_g = lrn("regr.rpart")
ml_m = lrn("classif.rpart")
obj_dml_data = make_irm_data(theta = 0.5)
dml_irm_obj = DoubleMLIRM$new(obj_dml_data, ml_g, ml_m)

param_grid = list(
  "ml_g" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)),
  "ml_m" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)))

# minimum requirements for tune_settings
tune_settings = list(
  terminator = mlr3tuning::trm("evals", n_evals = 5),
  algorithm = mlr3tuning::tnr("grid_search", resolution = 5))
dml_irm_obj$tune(param_set = param_grid, tune_settings = tune_settings)
dml_irm_obj$fit()
dml_irm_obj$summary()

## End(Not run)

Description

Double machine learning for partially linear IV regression models.

Format

R6::R6Class object inheriting from DoubleML.

Details

Partially linear IV regression (PLIV) models take the form

$Y - D\theta_0 = g_0(X) + \zeta$,

$Z = m_0(X) + V$,

with $E[\zeta|Z,X]=0$ and $E[V|X] = 0$. $Y$ is the outcome variable variable, $D$ is the policy variable of interest and $Z$ denotes one or multiple instrumental variables. The high-dimensional vector $X = (X_1, \ldots, X_p)$ consists of other confounding covariates, and $\zeta$ and $V$ are stochastic errors.

Super class

DoubleML::DoubleML -> DoubleMLPLIV

Active bindings

Methods

Public methods

• DoubleMLPLIV$new()

• DoubleMLPLIV$set_ml_nuisance_params()

• DoubleMLPLIV$tune()

• DoubleMLPLIV$clone()

Method new()

Creates a new instance of this R6 class.

Usage

DoubleMLPLIV$new(
  data,
  ml_l,
  ml_m,
  ml_r,
  ml_g = NULL,
  partialX = TRUE,
  partialZ = FALSE,
  n_folds = 5,
  n_rep = 1,
  score = "partialling out",
  dml_procedure = "dml2",
  draw_sample_splitting = TRUE,
  apply_cross_fitting = TRUE
)

Arguments

Method set_ml_nuisance_params()

Set hyperparameters for the nuisance models of DoubleML models.

Note that in the current implementation, either all parameters have to be set globally or all parameters have to be provided fold-specific.

Usage

DoubleMLPLIV$set_ml_nuisance_params(
  learner = NULL,
  treat_var = NULL,
  params,
  set_fold_specific = FALSE
)

Arguments

Returns

self

Method tune()

Hyperparameter-tuning for DoubleML models.

The hyperparameter-tuning is performed using the tuning methods provided in the mlr3tuning package. For more information on tuning in mlr3, we refer to the section on parameter tuning in the mlr3 book.

Usage

DoubleMLPLIV$tune(
  param_set,
  tune_settings = list(n_folds_tune = 5, rsmp_tune = mlr3::rsmp("cv", folds = 5), measure
    = NULL, terminator = mlr3tuning::trm("evals", n_evals = 20), algorithm =
    mlr3tuning::tnr("grid_search"), resolution = 5),
  tune_on_folds = FALSE
)

Arguments

Returns

self

Method clone()

The objects of this class are cloneable with this method.

Usage

DoubleMLPLIV$clone(deep = FALSE)

Arguments

See Also

Other DoubleML: DoubleML, DoubleMLIIVM, DoubleMLIRM, DoubleMLPLR

Examples

library(DoubleML)
library(mlr3)
library(mlr3learners)
library(data.table)
set.seed(2)
ml_l = lrn("regr.ranger", num.trees = 100, mtry = 20, min.node.size = 2, max.depth = 5)
ml_m = ml_l$clone()
ml_r = ml_l$clone()
obj_dml_data = make_pliv_CHS2015(alpha = 1, n_obs = 500, dim_x = 20, dim_z = 1)
dml_pliv_obj = DoubleMLPLIV$new(obj_dml_data, ml_l, ml_m, ml_r)
dml_pliv_obj$fit()
dml_pliv_obj$summary()


## Not run: 
library(DoubleML)
library(mlr3)
library(mlr3learners)
library(mlr3tuning)
library(data.table)
set.seed(2)
ml_l = lrn("regr.rpart")
ml_m = ml_l$clone()
ml_r = ml_l$clone()
obj_dml_data = make_pliv_CHS2015(
  alpha = 1, n_obs = 500, dim_x = 20,
  dim_z = 1)
dml_pliv_obj = DoubleMLPLIV$new(obj_dml_data, ml_l, ml_m, ml_r)
param_grid = list(
  "ml_l" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)),
  "ml_m" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)),
  "ml_r" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)))

# minimum requirements for tune_settings
tune_settings = list(
  terminator = mlr3tuning::trm("evals", n_evals = 5),
  algorithm = mlr3tuning::tnr("grid_search", resolution = 5))
dml_pliv_obj$tune(param_set = param_grid, tune_settings = tune_settings)
dml_pliv_obj$fit()
dml_pliv_obj$summary()

## End(Not run)

Description

Double machine learning for partially linear regression models.

Format

R6::R6Class object inheriting from DoubleML.

Details

Partially linear regression (PLR) models take the form

$Y = D\theta_0 + g_0(X) + \zeta,$

$D = m_0(X) + V,$

with $E[\zeta|D,X]=0$ and $E[V|X] = 0$. $Y$ is the outcome variable variable and $D$ is the policy variable of interest. The high-dimensional vector $X = (X_1, \ldots, X_p)$ consists of other confounding covariates, and $\zeta$ and $V$ are stochastic errors.

Super class

DoubleML::DoubleML -> DoubleMLPLR

Methods

Public methods

• DoubleMLPLR$new()

• DoubleMLPLR$set_ml_nuisance_params()

• DoubleMLPLR$tune()

• DoubleMLPLR$clone()

Method new()

Creates a new instance of this R6 class.

Usage

DoubleMLPLR$new(
  data,
  ml_l,
  ml_m,
  ml_g = NULL,
  n_folds = 5,
  n_rep = 1,
  score = "partialling out",
  dml_procedure = "dml2",
  draw_sample_splitting = TRUE,
  apply_cross_fitting = TRUE
)

Arguments

Method set_ml_nuisance_params()

Set hyperparameters for the nuisance models of DoubleML models.

Note that in the current implementation, either all parameters have to be set globally or all parameters have to be provided fold-specific.

Usage

DoubleMLPLR$set_ml_nuisance_params(
  learner = NULL,
  treat_var = NULL,
  params,
  set_fold_specific = FALSE
)

Arguments

Returns

self

Method tune()

Hyperparameter-tuning for DoubleML models.

The hyperparameter-tuning is performed using the tuning methods provided in the mlr3tuning package. For more information on tuning in mlr3, we refer to the section on parameter tuning in the mlr3 book.

Usage

DoubleMLPLR$tune(
  param_set,
  tune_settings = list(n_folds_tune = 5, rsmp_tune = mlr3::rsmp("cv", folds = 5), measure
    = NULL, terminator = mlr3tuning::trm("evals", n_evals = 20), algorithm =
    mlr3tuning::tnr("grid_search"), resolution = 5),
  tune_on_folds = FALSE
)

Arguments

Returns

self

Method clone()

The objects of this class are cloneable with this method.

Usage

DoubleMLPLR$clone(deep = FALSE)

Arguments

See Also

Other DoubleML: DoubleML, DoubleMLIIVM, DoubleMLIRM, DoubleMLPLIV

Examples

library(DoubleML)
library(mlr3)
library(mlr3learners)
library(data.table)
set.seed(2)
ml_g = lrn("regr.ranger", num.trees = 10, max.depth = 2)
ml_m = ml_g$clone()
obj_dml_data = make_plr_CCDDHNR2018(alpha = 0.5)
dml_plr_obj = DoubleMLPLR$new(obj_dml_data, ml_g, ml_m)
dml_plr_obj$fit()
dml_plr_obj$summary()


## Not run: 
library(DoubleML)
library(mlr3)
library(mlr3learners)
library(mlr3tuning)
library(data.table)
set.seed(2)
ml_l = lrn("regr.rpart")
ml_m = ml_l$clone()
obj_dml_data = make_plr_CCDDHNR2018(alpha = 0.5)
dml_plr_obj = DoubleMLPLR$new(obj_dml_data, ml_l, ml_m)

param_grid = list(
  "ml_l" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)),
  "ml_m" = paradox::ps(
    cp = paradox::p_dbl(lower = 0.01, upper = 0.02),
    minsplit = paradox::p_int(lower = 1, upper = 2)))

# minimum requirements for tune_settings
tune_settings = list(
  terminator = mlr3tuning::trm("evals", n_evals = 5),
  algorithm = mlr3tuning::tnr("grid_search", resolution = 5))
dml_plr_obj$tune(param_set = param_grid, tune_settings = tune_settings)
dml_plr_obj$fit()
dml_plr_obj$summary()

## End(Not run)

Description

Preprocessed data set on financial wealth and 401(k) plan participation. The raw data files are preprocessed to reproduce the examples in Chernozhukov et al. (2020). An internet connection is required to sucessfully download the data set.

Usage

fetch_401k(
  return_type = "DoubleMLData",
  polynomial_features = FALSE,
  instrument = FALSE
)

Arguments

Details

Variable description, based on the supplementary material of Chernozhukov et al. (2020):

• net_tfa: net total financial assets

• e401: = 1 if employer offers 401(k)

• p401: = 1 if individual participates in a 401(k) plan

• age: age

• inc: income

• fsize: family size

• educ: years of education

• db: = 1 if individual has defined benefit pension

• marr: = 1 if married

• twoearn: = 1 if two-earner household

• pira: = 1 if individual participates in IRA plan

• hown: = 1 if home owner

The supplementary data of the study by Chernozhukov et al. (2018) is available at https://academic.oup.com/ectj/article/21/1/C1/5056401#supplementary-data.

Value

A data object according to the choice of return_type.

References

Abadie, A. (2003), Semiparametric instrumental variable estimation of treatment response models. Journal of Econometrics, 113(2): 231-263.

Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and Robins, J. (2018), Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21: C1-C68. doi:10.1111/ectj.12097.

Description

Preprocessed data set on the Pennsylvania Reemploymnent Bonus experiment. The raw data files are preprocessed to reproduce the examples in Chernozhukov et al. (2020). An internet connection is required to sucessfully download the data set.

Usage

fetch_bonus(return_type = "DoubleMLData", polynomial_features = FALSE)

Arguments

Details

Variable description, based on the supplementary material of Chernozhukov et al. (2020):

• abdt: chronological time of enrollment of each claimant in the Pennsylvania reemployment bonus experiment.

• tg: indicates the treatment group (bonus amount - qualification period) of each claimant.

• inuidur1: a measure of length (in weeks) of the first spell of unemployment

• inuidur2: a second measure for the length (in weeks) of

• female: dummy variable; it indicates if the claimant's sex is female (=1) or male (=0).

• black: dummy variable; it indicates a person of black race (=1).

• hispanic: dummy variable; it indicates a person of hispanic race (=1).

• othrace: dummy variable; it indicates a non-white, non-black, not-hispanic person (=1).

• dep1: dummy variable; indicates if the number of dependents of each claimant is equal to 1 (=1).

• dep2: dummy variable; indicates if the number of dependents of each claimant is equal to 2 (=1).

• q1-q6: six dummy variables indicating the quarter of experiment during which each claimant enrolled.

• recall: takes the value of 1 if the claimant answered “yes” when was asked if he/she had any expectation to be recalled

• agelt35: takes the value of 1 if the claimant's age is less than 35 and 0 otherwise.

• agegt54: takes the value of 1 if the claimant's age is more than 54 and 0 otherwise.

• durable: it takes the value of 1 if the occupation of the claimant was in the sector of durable manufacturing and 0 otherwise.

• nondurable: it takes the value of 1 if the occupation of the claimant was in the sector of nondurable manufacturing and 0 otherwise.

• lusd: it takes the value of 1 if the claimant filed in Coatesville, Reading, or Lancaster and 0 otherwise.

• These three sites were considered to be located in areas characterized by low unemployment rate and short duration of unemployment.

• husd: it takes the value of 1 if the claimant filed in Lewistown, Pittston, or Scranton and 0 otherwise.

• These three sites were considered to be located in areas characterized by high unemployment rate and short duration of unemployment.

• muld: it takes the value of 1 if the claimant filed in Philadelphia-North, Philadelphia-Uptown, McKeesport, Erie, or Butler and 0 otherwise.

• These three sites were considered to be located in areas characterized by moderate unemployment rate and long duration of unemployment."

The supplementary data of the study by Chernozhukov et al. (2018) is available at https://academic.oup.com/ectj/article/21/1/C1/5056401#supplementary-data.

The supplementary data of the study by Bilias (2000) is available at https://www.journaldata.zbw.eu/dataset/sequential-testing-of-duration-data-the-case-of-the-pennsylvania-reemployment-bonus-experiment.

Value

A data object according to the choice of return_type.

References

Bilias Y. (2000), Sequential Testing of Duration Data: The Case of Pennsylvania ‘Reemployment Bonus’ Experiment. Journal of Applied Econometrics, 15(6): 575-594.

Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and Robins, J. (2018), Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21: C1-C68. doi:10.1111/ectj.12097.

Examples

library(DoubleML)
df_bonus = fetch_bonus(return_type = "data.table")
obj_dml_data_bonus = DoubleMLData$new(df_bonus,
  y_col = "inuidur1",
  d_cols = "tg",
  x_cols = c(
    "female", "black", "othrace", "dep1", "dep2",
    "q2", "q3", "q4", "q5", "q6", "agelt35", "agegt54",
    "durable", "lusd", "husd"
  )
)
obj_dml_data_bonus

Description

Generates data from a interactive IV regression (IIVM) model. The data generating process is defined as

$d_i = 1\left\lbrace \alpha_x Z + v_i > 0 \right\rbrace,$

$y_i = \theta d_i + x_i' \beta + u_i,$

$Z \sim \textstyle{Bernoulli} (0.5)$ and

$\left(\begin{array}{c} u_i \\ v_i \end{array} \right) \sim \mathcal{N}\left(0, \left(\begin{array}{cc} 1 & 0.3 \\ 0.3 & 1 \end{array} \right) \right).$

The covariates :$x_i \sim \mathcal{N}(0, \Sigma)$, where $\Sigma$ is a matrix with entries $\Sigma_{kj} = 0.5^{|j-k|}$ and $\beta$ is a dim_x-vector with entries $\beta_j=\frac{1}{j^2}$.

The data generating process is inspired by a process used in the simulation experiment of Farbmacher, Gruber and Klaaßen (2020).

Usage

make_iivm_data(
  n_obs = 500,
  dim_x = 20,
  theta = 1,
  alpha_x = 0.2,
  return_type = "DoubleMLData"
)

Arguments

References

Farbmacher, H., Guber, R. and Klaaßen, S. (2020). Instrument Validity Tests with Causal Forests. MEA Discussion Paper No. 13-2020. Available at SSRN:doi:10.2139/ssrn.3619201.

Description

Generates data from a interactive regression (IRM) model. The data generating process is defined as

$d_i = 1\left\lbrace \frac{\exp(c_d x_i' \beta)}{1+\exp(c_d x_i' \beta)} > v_i \right\rbrace,$

$y_i = \theta d_i + c_y x_i' \beta d_i + \zeta_i,$

with $v_i \sim \mathcal{U}(0,1)$, $\zeta_i \sim \mathcal{N}(0,1)$ and covariates $x_i \sim \mathcal{N}(0, \Sigma)$, where $\Sigma$ is a matrix with entries $\Sigma_{kj} = 0.5^{|j-k|}$. $\beta$ is a dim_x-vector with entries $\beta_j = \frac{1}{j^2}$ and the constancts $c_y$ and $c_d$ are given by

$c_y = \sqrt{\frac{R_y^2}{(1-R_y^2) \beta' \Sigma \beta}},$

$c_d = \sqrt{\frac{(\pi^2 /3) R_d^2}{(1-R_d^2) \beta' \Sigma \beta}}.$

The data generating process is inspired by a process used in the simulation experiment (see Appendix P) of Belloni et al. (2017).

Usage

make_irm_data(
  n_obs = 500,
  dim_x = 20,
  theta = 0,
  R2_d = 0.5,
  R2_y = 0.5,
  return_type = "DoubleMLData"
)

Arguments

References

Belloni, A., Chernozhukov, V., Fernández-Val, I. and Hansen, C. (2017). Program Evaluation and Causal Inference With High-Dimensional Data. Econometrica, 85: 233-298.

Description

Generates data from a partially linear IV regression model used in Chernozhukov, Hansen and Spindler (2015). The data generating process is defined as

$z_i = \Pi x_i + \zeta_i,$

$d_i = x_i'\gamma + z_i'\delta + u_i,$

$y_i = \alpha d_i + x_i'\beta + \epsilon_i,$

with

$\left(\begin{array}{c} \varepsilon_i \\ u_i \\ \zeta_i \\ x_i \end{array} \right) \sim \mathcal{N}\left(0, \left(\begin{array}{cccc} 1 & 0.6 & 0 & 0 \\ 0.6 & 1 & 0 & 0 \\ 0 & 0 & 0.25 I_{p_n^z} & 0 \\ 0 & 0 & 0 & \Sigma \end{array} \right) \right)$

where $\Sigma$ is a $p_n^x \times p_n^x$ matrix with entries $\Sigma_{kj} = 0.5^{|j-k|}$ and $I_{p_n^z}$ is the $p^z_n \times p^z_n$ identity matrix. $\beta=\gamma$ iis a $p^x_n$-vector with entries $\beta_j = \frac{1}{j^2}$, $\delta$ is a $p^z_n$-vector with entries $\delta_j = \frac{1}{j^2}$ and $\Pi = (I_{p_n^z}, O_{p_n^z \times (p_n^x - p_n^z)})$.

Usage

make_pliv_CHS2015(
  n_obs,
  alpha = 1,
  dim_x = 200,
  dim_z = 150,
  return_type = "DoubleMLData"
)

Arguments

Value

A data object according to the choice of return_type.

References

Chernozhukov, V., Hansen, C. and Spindler, M. (2015), Post-Selection and Post-Regularization Inference in Linear Models with Many Controls and Instruments. American Economic Review: Papers and Proceedings, 105 (5): 486-90.

Description

Generates data from a partially linear IV regression model with multiway cluster sample used in Chiang et al. (2021). The data generating process is defined as

$Z_{ij} = X_{ij}' \xi_0 + V_{ij},$

$D_{ij} = Z_{ij}' \pi_{10} + X_{ij}' \pi_{20} + v_{ij},$

$Y_{ij} = D_{ij} \theta + X_{ij}' \zeta_0 + \varepsilon_{ij},$

with

$X_{ij} = (1 - \omega_1^X - \omega_2^X) \alpha_{ij}^X + \omega_1^X \alpha_{i}^X + \omega_2^X \alpha_{j}^X,$

$\varepsilon_{ij} = (1 - \omega_1^\varepsilon - \omega_2^\varepsilon) \alpha_{ij}^\varepsilon + \omega_1^\varepsilon \alpha_{i}^\varepsilon + \omega_2^\varepsilon \alpha_{j}^\varepsilon,$

$v_{ij} = (1 - \omega_1^v - \omega_2^v) \alpha_{ij}^v + \omega_1^v \alpha_{i}^v + \omega_2^v \alpha_{j}^v,$

$V_{ij} = (1 - \omega_1^V - \omega_2^V) \alpha_{ij}^V + \omega_1^V \alpha_{i}^V + \omega_2^V \alpha_{j}^V,$

and $\alpha_{ij}^X, \alpha_{i}^X, \alpha_{j}^X \sim \mathcal{N}(0, \Sigma)$ where $\Sigma$ is a $p_x \times p_x$ matrix with entries $\Sigma_{kj} = s_X^{|j-k|}$.

Further

$\left(\begin{array}{c} \alpha_{ij}^\varepsilon \\ \alpha_{ij}^v \end{array}\right), \left(\begin{array}{c} \alpha_{i}^\varepsilon \\ \alpha_{i}^v \end{array}\right), \left(\begin{array}{c} \alpha_{j}^\varepsilon \\ \alpha_{j}^v \end{array}\right) \sim \mathcal{N}\left(0, \left(\begin{array}{cc} 1 & s_{\varepsilon v} \\ s_{\varepsilon v} & 1 \end{array}\right) \right)$

and $\alpha_{ij}^V, \alpha_{i}^V, \alpha_{j}^V \sim \mathcal{N}(0, 1)$.

Usage

make_pliv_multiway_cluster_CKMS2021(
  N = 25,
  M = 25,
  dim_X = 100,
  theta = 1,
  return_type = "DoubleMLClusterData",
  ...
)

Arguments

Value

A data object according to the choice of return_type.

References

Chiang, H. D., Kato K., Ma, Y. and Sasaki, Y. (2021), Multiway Cluster Robust Double/Debiased Machine Learning, Journal of Business & Economic Statistics, doi:10.1080/07350015.2021.1895815, https://arxiv.org/abs/1909.03489.

Description

Generates data from a partially linear regression model used in Chernozhukov et al. (2018) for Figure 1. The data generating process is defined as

$d_i = m_0(x_i) + s_1 v_i,$

$y_i = \alpha d_i + g_0(x_i) + s_2 \zeta_i,$

with $v_i \sim \mathcal{N}(0,1)$ and $\zeta_i \sim \mathcal{N}(0,1),$. The covariates are distributed as $x_i \sim \mathcal{N}(0, \Sigma)$, where $\Sigma$ is a matrix with entries $\Sigma_{kj} = 0.7^{|j-k|}$. The nuisance functions are given by

$m_0(x_i) = a_0 x_{i,1} + a_1 \frac{\exp(x_{i,3})}{1+\exp(x_{i,3})},$

$g_0(x_i) = b_0 \frac{\exp(x_{i,1})}{1+\exp(x_{i,1})} + b_1 x_{i,3},$

with $a_0=1$, $a_1=0.25$, $s_1=1$, $b_0=1$, $b_1=0.25$, $s_2=1$.

Usage

make_plr_CCDDHNR2018(
  n_obs = 500,
  dim_x = 20,
  alpha = 0.5,
  return_type = "DoubleMLData"
)

Arguments

Value

A data object according to the choice of return_type.

References

Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and Robins, J. (2018), Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21: C1-C68. doi:10.1111/ectj.12097.

Description

Generates data from a partially linear regression model used in a blog article by Turrell (2018). The data generating process is defined as

$d_i = m_0(x_i' b) + v_i,$

$y_i = \theta d_i + g_0(x_i' b) + u_i,$

with $v_i \sim \mathcal{N}(0,1)$, $u_i \sim \mathcal{N}(0,1)$, and covariates $x_i \sim \mathcal{N}(0, \Sigma)$, where $\Sigma$ is a random symmetric, positive-definite matrix generated with clusterGeneration::genPositiveDefMat(). $b$ is a vector with entries $b_j=\frac{1}{j}$ and the nuisance functions are given by

$m_0(x_i) = \frac{1}{2 \pi} \frac{\sinh(\gamma)}{\cosh(\gamma) - \cos(x_i-\nu)},$

$g_0(x_i) = \sin(x_i)^2.$

Usage

make_plr_turrell2018(
  n_obs = 100,
  dim_x = 20,
  theta = 0.5,
  return_type = "DoubleMLData",
  nu = 0,
  gamma = 1
)

Arguments