Type: Package
Title: Causal Inference with Super Learner and Deep Neural Networks
Version: 0.0.106
Maintainer: Nguyen K. Huynh <khoinguyen.huynh@r.hit-u.ac.jp>
Description: Functions to estimate Conditional Average Treatment Effects (CATE) and Population Average Treatment Effects on the Treated (PATT) from experimental or observational data using the Super Learner (SL) ensemble method and Deep neural networks. The package first provides functions to implement meta-learners such as the Single-learner (S-learner) and Two-learner (T-learner) described in KC<nzel et al. (2019) <doi:10.1073/pnas.1804597116> for estimating the CATE. The S- and T-learner are each estimated using the SL ensemble method and deep neural networks. It then provides functions to implement the Ottoboni and Poulos (2020) <doi:10.1515/jci-2018-0035> PATT-C estimator to obtain the PATT from experimental data with noncompliance by using the SL ensemble method and deep neural networks.
License: GPL-3
Encoding: UTF-8
LazyData: true
Imports: ROCR, caret, neuralnet, SuperLearner, class, xgboost, randomForest, glmnet, gam, e1071, gbm, Hmisc, ggplot2, dplyr, tidyr, magrittr, weights
Suggests: testthat,
RoxygenNote: 7.3.2
Depends: R (≥ 4.1.0)
URL: https://github.com/hknd23/DeepLearningCausal
BugReports: https://github.com/hknd23/DeepLearningCausal/issues
NeedsCompilation: no
Packaged: 2025-06-11 14:16:06 UTC; nguye
Author: Nguyen K. Huynh ORCID iD [aut, cre], Bumba Mukherjee ORCID iD [aut], Yang Yang ORCID iD [aut]
Repository: CRAN
Date/Publication: 2025-06-11 16:00:05 UTC

Train complier model using ensemble methods

Description

Train model using group exposed to treatment with compliance as binary outcome variable and covariates.

Usage

complier_mod(
  exp.data,
  complier.formula,
  treat.var,
  ID = NULL,
  SL.learners = c("SL.glmnet", "SL.xgboost", "SL.ranger", "SL.nnet", "SL.glm")
)

Arguments

exp.data

list object of experimental data.

complier.formula

formula to fit compliance model (c ~ x) using complier variable and covariates

treat.var

string specifying the binary treatment variable

ID

string for name of indentifier variable.

SL.learners

vector of strings for ML classifier algorithms. If left NULL employs extreme gradient boosting, elastic net regression, random forest, and neural nets.

Value

model object of trained model.

Complier model prediction

Description

Predict Compliance from control group in experimental data

Usage

complier_predict(complier.mod, exp.data, treat.var, compl.var)

Arguments

complier.mod

output from trained ensemble superlearner model

exp.data

data.frame object of experimental dataset

treat.var

string specifying the binary treatment variable

compl.var

string specifying binary complier variable

Value

data.frame object with true compliers, predicted compliers in the control group, and all compliers (actual + predicted).

Survey Experiment of Support for Populist Policy

Description

Shortened version of survey response data that incorporates a vignette survey experiment. The vignette describes an international crisis between country A and B. After reading this vignette, respondents are randomly assigned to the control group or to one of two treatments: policy prescription to said crisis by strong (populist) leader and centrist (non-populist) leader. The respondents are then asked whether they are willing to support the policy decision to fight a war against country A, which is the dependent variable.

Usage

data(exp_data)

Format

exp_data

A data frame with 257 rows and 12 columns:

female

Gender.

age

Age of participant.

income

Monthly household income.

religion

Religious denomination

practicing_religion

Importance of religion in life.

education

Educational level of participant.

political_ideology

Political ideology of participant.

employment

Employment status of participant.

marital_status

Marital status of participant.

job_loss

Concern about job loss.

strong_leader

Binary treatment measure of leader type.

support_war

Binary outcome measure for willingness to fight war.

#' ...

Source

Yadav and Mukherjee (2024)

Survey Experiment of Support for Populist Policy

Description

Extended experiment data with 514 observations

Usage

data(exp_data_full)

Format

exp_data_full

A data frame with 514 rows and 12 columns:

female

Gender.

age

Age of participant.

income

Monthly household income.

religion

Religious denomination

practicing_religion

Importance of religion in life.

education

Educational level of participant.

political_ideology

Political ideology of participant.

employment

Employment status of participant.

marital_status

Marital status of participant.

job_loss

Concern about job loss.

strong_leader

Binary treatment measure of leader type.

support_war

Binary outcome measure for willingness to fight war.

#' ...

Source

Yadav and Mukherjee (2024)

Create list for experimental data

Description

create list object of experimental data for easy data processing

Usage

expcall(
  response.formula,
  treat.var,
  compl.var,
  exp.data,
  weights = NULL,
  cluster = NULL,
  ID = NULL
)

Arguments

response.formula

formula for response equation of binary outcome variable and covariates

treat.var

string for binary treatment variable

compl.var

string for complier variable

exp.data

data.frame of experimental variable

weights

observation weights

cluster

clustering variable

ID

identifier variable

Value

list of processed dataset

hte_plot

Description

Produces plot to illustrate sub-group Heterogeneous Treatment Effects (HTE) of estimated CATEs from metalearner_ensemble and metalearner_deepneural, as well as PATT-C from pattc_ensemble and pattc_neural.

Usage

hte_plot(
  x,
  ...,
  boot = TRUE,
  n_boot = 1000,
  cut_points = NULL,
  custom_labels = NULL,
  zero_int = TRUE
)

Arguments

x

estimated model from metalearner_ensemble, metalearner_deepneural, pattc_ensemble, or pattc_neural.

...

Additional arguments

boot

logical for using bootstraps to estimate confidence intervals.

n_boot

number of bootstrap iterations. Only used with boot = TRUE.

cut_points

numeric vector for cut-off points to generate subgroups from covariates. If left blank a vector generated from median values will be used.

custom_labels

character vector for the names of subgroups.

zero_int

logical for vertical line at 0 x intercept.

Value

ggplot object illustrating subgroup HTE and 95% confidence intervals.

Examples


# load dataset
set.seed(123456)
xlearner_nn <- metalearner_deepneural(cov.formula = support_war ~ age +
                                  income  + employed  + job_loss,
                                  data = exp_data,
                                  treat.var = "strong_leader",
                                  meta.learner.type = "X.Learner",
                                  stepmax = 2e+9,
                                  nfolds = 5,
                                  algorithm = "rprop+",
                                  hidden.layer = c(3),
                                  linear.output = FALSE,
                                  binary.outcome = FALSE)

hte_plot(xlearner_nn)

metalearner_deepneural

Description

metalearner_deepneural implements the S-learner and T-learner for estimating CATE using Deep Neural Networks. The Resilient back propagation (Rprop) algorithm is used for training neural networks.

Usage

metalearner_deepneural(
  data,
  cov.formula,
  treat.var,
  meta.learner.type,
  stepmax = 1e+05,
  nfolds = 5,
  algorithm = "rprop+",
  hidden.layer = c(4, 2),
  linear.output = FALSE,
  binary.outcome = FALSE
)

Arguments

data

data.frame object of data.

cov.formula

formula description of the model y ~ x(list of covariates).

treat.var

string for the name of treatment variable.

meta.learner.type

string specifying is the S-learner and "T.Learner" for the T-learner model.

stepmax

maximum number of steps for training model.

nfolds

number of folds for cross-validation. Currently supports up to 5 folds.

algorithm

a string for the algorithm for the neural network. Default set to ⁠rprop+⁠, the Resilient back propagation (Rprop) with weight backtracking algorithm for training neural networks.

hidden.layer

vector of integers specifying layers and number of neurons.

linear.output

logical specifying regression (TRUE) or classification (FALSE) model.

binary.outcome

logical specifying predicted outcome variable will take binary values or proportions.

Value

metalearner_deepneural of predicted outcome values and CATEs estimated by the meta learners for each observation.

Examples


# load dataset
data(exp_data)
# estimate CATEs with S Learner
set.seed(123456)
slearner_nn <- metalearner_deepneural(cov.formula = support_war ~ age + income +
                                   employed  + job_loss,
                                   data = exp_data,
                                   treat.var = "strong_leader",
                                   meta.learner.type = "S.Learner",
                                   stepmax = 2e+9,
                                   nfolds = 5,
                                   algorithm = "rprop+",
                                   hidden.layer = c(1),
                                   linear.output = FALSE,
                                   binary.outcome = FALSE)

print(slearner_nn)

# load dataset
set.seed(123456)
# estimate CATEs with T Learner
tlearner_nn <- metalearner_deepneural(cov.formula = support_war ~ age +
                                  income  +
                                  employed  + job_loss,
                                  data = exp_data,
                                  treat.var = "strong_leader",
                                  meta.learner.type = "T.Learner",
                                  stepmax = 1e+9,
                                  nfolds = 5,
                                  algorithm = "rprop+",
                                  hidden.layer = c(2,1),
                                  linear.output = FALSE,
                                  binary.outcome = FALSE)

print(tlearner_nn)

# load dataset
set.seed(123456)
# estimate CATEs with X Learner
xlearner_nn <- metalearner_deepneural(cov.formula = support_war ~ age +
                                  income  +
                                  employed  + job_loss,
                                  data = exp_data,
                                  treat.var = "strong_leader",
                                  meta.learner.type = "X.Learner",
                                  stepmax = 2e+9,
                                  nfolds = 5,
                                  algorithm = "rprop+",
                                  hidden.layer = c(3),
                                  linear.output = FALSE,
                                  binary.outcome = FALSE)

print(xlearner_nn)

metalearner_ensemble

Description

metalearner_ensemble implements the S-learner, T-learner, and X-learner for estimating CATE using the super learner ensemble method. The super learner in this case includes the following machine learning algorithms: extreme gradient boosting, glmnet (elastic net regression), random forest and neural nets.

Usage

metalearner_ensemble(
  data,
  cov.formula,
  treat.var,
  meta.learner.type,
  SL.learners = c("SL.glmnet", "SL.xgboost", "SL.ranger", "SL.nnet"),
  nfolds = 5,
  binary.outcome = FALSE
)

Arguments

data

data.frame object of data

cov.formula

formula description of the model y ~ x(list of covariates)

treat.var

string for the name of treatment variable.

meta.learner.type

string specifying is the S-learner and "T.Learner" for the T-learner model. "X.Learner" for the X-learner model.

SL.learners

vector for super learner ensemble that includes extreme gradient boosting, glmnet, random forest, and neural nets.

nfolds

number of folds for cross-validation. Currently supports up to 5 folds.

binary.outcome

logical specifying predicted outcome variable will take binary values or proportions.

Value

metalearner_ensemble of predicted outcome values and CATEs estimated by the meta learners for each observation.

Examples

# load dataset
data(exp_data)
#load SuperLearner package
library(SuperLearner)
# estimate CATEs with S Learner
set.seed(123456)
slearner <- metalearner_ensemble(cov.formula = support_war ~ age +
                                  income + employed + job_loss,
                                data = exp_data,
                                treat.var = "strong_leader",
                                meta.learner.type = "S.Learner",
                                SL.learners = c("SL.glm"),
                                nfolds = 5,
                                binary.outcome = FALSE)
print(slearner)


# estimate CATEs with T Learner
set.seed(123456)
tlearner <- metalearner_ensemble(cov.formula = support_war ~ age + income +
                                  employed  + job_loss,
                                  data = exp_data,
                                  treat.var = "strong_leader",
                                  meta.learner.type = "T.Learner",
                                  SL.learners = c("SL.xgboost","SL.ranger",
                                               "SL.nnet"),
                                  nfolds = 5,
                                  binary.outcome = FALSE)

print(tlearner)
                                  


# estimate CATEs with X Learner
set.seed(123456)
xlearner <- metalearner_ensemble(cov.formula = support_war ~ age + income +
 employed  + job_loss,
                                 data = exp_data,
                                 treat.var = "strong_leader",
                                 meta.learner.type = "X.Learner",
                                 SL.learners = c("SL.glmnet","SL.xgboost", 
                                 "SL.ranger","SL.nnet"),
                                 nfolds = 5,
                                 binary.outcome = TRUE)

print(xlearner)

Train compliance model using neural networks

Description

Train model using group exposed to treatment with compliance as binary outcome variable and covariates.

Usage

neuralnet_complier_mod(
  complier.formula,
  exp.data,
  treat.var,
  algorithm = "rprop+",
  hidden.layer = c(4, 2),
  ID = NULL,
  stepmax = 1e+08
)

Arguments

complier.formula

formula for complier variable as outcome and covariates (c ~ x)

exp.data

data.frame for experimental data.

treat.var

string for treatment variable.

algorithm

string for algorithm for training neural networks. Default set to the Resilient back propagation with weight backtracking (rprop+). Other algorithms include backprop', rprop-', 'sag', or 'slr' (see neuralnet package).

hidden.layer

vector for specifying hidden layers and number of neurons.

ID

string for identifier variable

stepmax

maximum number of steps.

Value

trained complier model object

Assess Population Data counterfactuals

Description

Create counterfactual datasets in the population for compliers and noncompliers. Then predict potential outcomes using trained model from neuralnet_response_model.

Usage

neuralnet_pattc_counterfactuals(
  pop.data,
  neuralnet.response.mod,
  ID = NULL,
  cluster = NULL,
  binary.outcome = FALSE
)

Arguments

pop.data

population data.

neuralnet.response.mod

trained model from. neuralnet_response_model.

ID

string for identifier variable.

cluster

string for clustering variable (currently unused).

binary.outcome

logical specifying predicted outcome variable will take binary values or proportions.

Value

data.frame of predicted outcomes of response variable from counterfactuals.

Predicting Compliance from experimental data

Description

Predicting Compliance from control group experimental data

Usage

neuralnet_predict(neuralnet.complier.mod, exp.data, treat.var, compl.var)

Arguments

neuralnet.complier.mod

results from neuralnet_complier_mod

exp.data

data.frame of experimental data

treat.var

string for treatment variable

compl.var

string for compliance variable

Value

data.frame object with true compliers, predicted compliers in the control group, and all compliers (actual + predicted).

Modeling Responses from experimental data Using Deep NN

Description

Model Responses from all compliers (actual + predicted) in experimental data using neural network.

Usage

neuralnet_response_model(
  response.formula,
  exp.data,
  neuralnet.compliers,
  compl.var,
  algorithm = "rprop+",
  hidden.layer = c(4, 2),
  stepmax = 1e+08
)

Arguments

response.formula

formula for response variable and covariates (y ~ x)

exp.data

data.frame of experimental data.

neuralnet.compliers

data.frame of compliers (actual + predicted) from neuralnet_predict.

compl.var

string of compliance variable

algorithm

neural network algorithm, default set to "rprop+".

hidden.layer

vector specifying hidden layers and number of neurons.

stepmax

maximum number of steps for training model.

Value

trained response model object

Assess Population Data counterfactuals

Description

Create counterfactual datasets in the population for compliers and noncompliers. Then predict potential outcomes from counterfactuals.

Usage

pattc_counterfactuals(
  pop.data,
  response.mod,
  ID = NULL,
  cluster = NULL,
  binary.outcome = FALSE
)

Arguments

pop.data

population dataset

response.mod

trained model from response_model.

ID

string fir identifier variable

cluster

string for clustering variable

binary.outcome

logical specifying whether predicted outcomes are proportions or binary (0-1).

Value

data.frame object of predicted outcomes of counterfactual groups.

Estimate PATT_C using Deep NN

Description

estimates the Population Average Treatment Effect of the Treated from experimental data with noncompliers using Deep Neural Networks.

Usage

pattc_deepneural(
  response.formula,
  exp.data,
  pop.data,
  treat.var,
  compl.var,
  compl.algorithm = "rprop+",
  response.algorithm = "rprop+",
  compl.hidden.layer = c(4, 2),
  response.hidden.layer = c(4, 2),
  compl.stepmax = 1e+08,
  response.stepmax = 1e+08,
  ID = NULL,
  cluster = NULL,
  binary.outcome = FALSE,
  bootstrap = FALSE,
  nboot = 1000
)

Arguments

response.formula

formula of response variable as outcome and covariates (y ~ x)

exp.data

data.frame of experimental data. Must include binary treatment and compliance variables.

pop.data

data.frame of population data. Must include binary compliance variable

treat.var

string for treatment variable.

compl.var

string for compliance variable

compl.algorithm

string for algorithim to train neural network for compliance model. Default set to "rprop+". See (neuralnet package for available algorithms).

response.algorithm

string for algorithim to train neural network for response model. Default set to "rprop+". See (neuralnet package for available algorithms).

compl.hidden.layer

vector for specifying hidden layers and number of neurons in complier model.

response.hidden.layer

vector for specifying hidden layers and number of neurons in response model.

compl.stepmax

maximum number of steps for complier model

response.stepmax

maximum number of steps for response model

ID

string for identifier variable

cluster

string for cluster variable.

binary.outcome

logical specifying predicted outcome variable will take binary values or proportions.

bootstrap

logical for bootstrapped PATT-C.

nboot

number of bootstrapped samples

Value

pattc_deepneural class object of results of t test as PATTC estimate.

Examples


# load datasets
data(exp_data) #experimental data
data(pop_data) #population data
# specify models and estimate PATTC
set.seed(123456)
pattc_neural <- pattc_deepneural(response.formula = support_war ~ age + female +
                               income + education +  employed + married +
                               hindu + job_loss,
                               exp.data = exp_data,
                               pop.data = pop_data,
                               treat.var = "strong_leader",
                               compl.var = "compliance",
                               compl.algorithm = "rprop+",
                               response.algorithm = "rprop+",
                               compl.hidden.layer = c(4,2),
                               response.hidden.layer = c(4,2),
                               compl.stepmax = 1e+09,
                               response.stepmax = 1e+09,
                               ID = NULL,
                               cluster = NULL,
                               binary.outcome = FALSE)

print(pattc_neural)

pattc_neural_boot <- pattc_deepneural(response.formula = support_war ~ age + female +
                               income + education +  employed + married +
                               hindu + job_loss,
                               exp.data = exp_data,
                               pop.data = pop_data,
                               treat.var = "strong_leader",
                               compl.var = "compliance",
                               compl.algorithm = "rprop+",
                               response.algorithm = "rprop+",
                               compl.hidden.layer = c(4,2),
                               response.hidden.layer = c(4,2),
                               compl.stepmax = 1e+09,
                               response.stepmax = 1e+09,
                               ID = NULL,
                               cluster = NULL,
                               binary.outcome = FALSE,
                               bootstrap = TRUE,
                               nboot = 2000)

print(pattc_neural_boot)

PATT_C SL Ensemble

Description

pattc_ensemble estimates the Population Average Treatment Effect of the Treated from experimental data with noncompliers using the super learner ensemble that includes extreme gradient boosting, glmnet (elastic net regression), random forest and neural nets.

Usage

pattc_ensemble(
  response.formula,
  exp.data,
  pop.data,
  treat.var,
  compl.var,
  compl.SL.learners = c("SL.glmnet", "SL.xgboost", "SL.ranger", "SL.nnet", "SL.glm"),
  response.SL.learners = c("SL.glmnet", "SL.xgboost", "SL.ranger", "SL.nnet", "SL.glm"),
  ID = NULL,
  cluster = NULL,
  binary.outcome = FALSE,
  bootstrap = FALSE,
  nboot = 1000
)

Arguments

response.formula

formula for the effects of covariates on outcome variable (y ~ x).

exp.data

data.frame object for experimental data. Must include binary treatment and compliance variable.

pop.data

data.frame object for population data. Must include binary compliance variable.

treat.var

string for binary treatment variable.

compl.var

string for binary compliance variable.

compl.SL.learners

vector of names of ML algorithms used for compliance model.

response.SL.learners

vector of names of ML algorithms used for response model.

ID

string for name of identifier. (currently not used)

cluster

string for name of cluster variable. (currently not used)

binary.outcome

logical specifying predicted outcome variable will take binary values or proportions.

bootstrap

logical for bootstrapped PATT-C.

nboot

number of bootstrapped samples. Only used with bootstrap = FALSE

Value

pattc_ensemble object of results of t test as PATTC estimate.

Examples


# load datasets
data(exp_data_full) # full experimental data
data(exp_data) #experimental data
data(pop_data) #population data
#attach SuperLearner (model will not recognize learner if package is not loaded)
library(SuperLearner)
set.seed(123456)
#specify models and estimate PATTC
pattc <- pattc_ensemble(response.formula = support_war ~ age + income +
                                education + employed + job_loss,
                                exp.data = exp_data_full,
                                pop.data = pop_data,
                                treat.var = "strong_leader",
                                compl.var = "compliance",
                                compl.SL.learners = c("SL.glm", "SL.nnet"),
                                response.SL.learners = c("SL.glm", "SL.nnet"),
                                ID = NULL,
                                cluster = NULL,
                                binary.outcome = FALSE)

print(pattc)

pattc_boot <- pattc_ensemble(response.formula = support_war ~ age + income +
                                education + employed + job_loss,
                                exp.data = exp_data_full,
                                pop.data = pop_data,
                                treat.var = "strong_leader",
                                compl.var = "compliance",
                                compl.SL.learners = c("SL.glm", "SL.nnet"),
                                response.SL.learners = c("SL.glm", "SL.nnet"),
                                ID = NULL,
                                cluster = NULL,
                                binary.outcome = FALSE,
                                bootstrap = TRUE,
                                nboot = 1000)
print(pattc_boot)

plot.metalearner_deepneural

Description

Uses plot() to generate histogram of ditribution of CATEs or predicted outcomes from metalearner_deepneural

Usage

## S3 method for class 'metalearner_deepneural'
plot(x, ..., conf_level = 0.95, type = "CATEs")

Arguments

x

metalearner_deepneural model object.

...

Additional arguments

conf_level

numeric value for confidence level. Defaults to 0.95.

type

"CATEs" or "predict".

Value

ggplot object.

plot.metalearner_ensemble

Description

Uses plot() to generate histogram of ditribution of CATEs or predicted outcomes from metalearner_ensemble

Usage

## S3 method for class 'metalearner_ensemble'
plot(x, ..., conf_level = 0.95, type = "CATEs")

Arguments

x

metalearner_ensemble model object

...

Additional arguments

conf_level

numeric value for confidence level. Defaults to 0.95.

type

"CATEs" or "predict"

Value

ggplot object

plot.pattc_deepneural

Description

Uses plot() to generate histogram of ditribution of CATEs or predicted outcomes from pattc_deepneural

Usage

## S3 method for class 'pattc_deepneural'
plot(x, ...)

Arguments

x

pattc_deepneural model object

...

Additional arguments

Value

ggplot object

plot.pattc_ensemble

Description

Uses plot() to generate histogram of ditribution of CATEs or predicted outcomes from pattc_ensemble

Usage

## S3 method for class 'pattc_ensemble'
plot(x, ...)

Arguments

x

pattc_ensemble model object

...

Additional arguments

Value

ggplot object

World Value Survey India Sample

Description

World Value Survey (WVS) Data for India in 2022. The variables drawn from the said WVS India data match the covariates from the India survey experiment sample.

Usage

data(pop_data)

Format

pop_data

A data frame with 846 rows and 13 columns:

female

Respondent’s Sex.

age

Age of respondent.

income

income group of Household.

religion

Religious denomination

practicing_religion

Importance of religion in respondent’s life.

education

Educational level of respondent.

political_ideology

Political ideology of respondent.

employment

Employment status and full-time employee.

marital_status

Marital status of respondent.

job_loss

Concern about job loss.

support_war

Binary (Yes/No) outcome measure for willingness to fight war.

strong_leader

Binary measure of preference for strong leader.

...

Source

Haerpfer, C., Inglehart, R., Moreno, A., Welzel, C., Kizilova, K., Diez-Medrano J., M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2020. World Values Survey: Round Seven – Country-Pooled Datafile. Madrid, Spain & Vienna, Austria: JD Systems Institute & WVSA Secretariat. <doi.org/10.14281/18241.1>

World Value Survey India Sample

Description

Extended World Value Survey (WVS) Data for India in 1995, 2001, 2006, 2012, and 2022.

Usage

data(pop_data_full)

Format

pop_data_full

A data frame with 11,813 rows and 13 columns:

female

Respondent’s Sex.

age

Age of respondent.

income

income group of Household.

religion

Religious denomination

practicing_religion

Importance of religion in respondent’s life.

education

Educational level of respondent.

political_ideology

Political ideology of respondent.

employment

Employment status and full-time employee.

marital_status

Marital status of respondent.

job_loss

Concern about job loss.

support_war

Binary (Yes/No) outcome measure for willingness to fight war.

strong_leader

Binary measure of preference for strong leader.

...

Source

Haerpfer, C., Inglehart, R., Moreno, A., Welzel, C., Kizilova, K., Diez-Medrano J., M. Lagos, P. Norris, E. Ponarin & B. Puranen et al. (eds.). 2020. World Values Survey: Round Seven – Country-Pooled Datafile. Madrid, Spain & Vienna, Austria: JD Systems Institute & WVSA Secretariat. <doi.org/10.14281/18241.1>

Create list for population data

Description

create list object of population data for easy data processing

Usage

popcall(
  response.formula,
  compl.var,
  treat.var,
  pop.data,
  weights = NULL,
  cluster = NULL,
  ID = NULL,
  patt = TRUE
)

Arguments

response.formula

formula for response equation of binary outcome variable and covariates

compl.var

string for complier variable

treat.var

string for treatmet variable

pop.data

data.frame of experimental variable

weights

observation weights

cluster

clustering variable

ID

identifier variable

patt

logical for patt, subsetting population treated observations

Value

list of processed dataset

print.metalearner_deepneural

Description

Print method for metalearner_deepneural

Usage

## S3 method for class 'metalearner_deepneural'
print(x, ...)

Arguments

x

metalearner_deepneural class object from metalearner_deepneural

...

additional parameter

Value

list of model results

print.metalearner_ensemble

Description

Print method for metalearner_ensemble

Usage

## S3 method for class 'metalearner_ensemble'
print(x, ...)

Arguments

x

metalearner_ensemble class object from metalearner_ensemble

...

additional parameter

Value

list of model results

print.pattc_deepneural

Description

Print method for pattc_deepneural

Usage

## S3 method for class 'pattc_deepneural'
print(x, ...)

Arguments

x

pattc_deepneural class object from pattc_deepneural

...

additional parameter

Value

list of model results

print.pattc_ensemble

Description

Print method for pattc_ensemble

Usage

## S3 method for class 'pattc_ensemble'
print(x, ...)

Arguments

x

pattc_ensemble class object from pattc_ensemble

...

additional parameter

Value

list of model results

Response model from experimental data using SL ensemble

Description

Train response model (response variable as outcome and covariates) from all compliers (actual + predicted) in experimental data using SL ensemble.

Usage

response_model(
  response.formula,
  exp.data,
  compl.var,
  exp.compliers,
  family = "binomial",
  ID = NULL,
  SL.learners = c("SL.glmnet", "SL.xgboost", "SL.ranger", "SL.nnet", "SL.glm")
)

Arguments

response.formula

formula to fit the response model (y ~ x) using binary outcome variable and covariates

exp.data

experimental dataset.

compl.var

string specifying binary complier variable

exp.compliers

data.frame object of compliers from complier_predict.

family

string for "gaussian" or "binomial".

ID

string for identifier variable.

SL.learners

vector of names of ML algorithms used for ensemble model.

Value

trained response model.