Title:	Augmented Inverse Probability Weighting
Version:	0.6.9.2
Maintainer:	Yongqi Zhong <yq.zhong7@gmail.com>
Description:	The 'AIPW' package implements the augmented inverse probability weighting, a doubly robust estimator, for average causal effect estimation with user-defined stacked machine learning algorithms. To cite the 'AIPW' package, please use: "Yongqi Zhong, Edward H. Kennedy, Lisa M. Bodnar, Ashley I. Naimi (2021). AIPW: An R Package for Augmented Inverse Probability Weighted Estimation of Average Causal Effects. American Journal of Epidemiology. <doi:10.1093/aje/kwab207>". Visit: https://yqzhong7.github.io/AIPW/ for more information.
License:	GPL-3
Encoding:	UTF-8
Language:	es
LazyData:	true
Suggests:	testthat (≥ 2.1.0), knitr, rmarkdown, covr, tmle
RoxygenNote:	7.2.2
Imports:	stats, utils, R6, SuperLearner, ggplot2, future.apply, progressr, Rsolnp
URL:	https://github.com/yqzhong7/AIPW
BugReports:	https://github.com/yqzhong7/AIPW/issues
VignetteBuilder:	knitr
Depends:	R (≥ 2.10)
NeedsCompilation:	no
Packaged:	2025-04-05 16:43:11 UTC; k
Author:	Yongqi Zhong [aut, cre], Ashley Naimi [aut], Gabriel Conzuelo [ctb], Edward Kennedy [ctb]
Repository:	CRAN
Date/Publication:	2025-04-05 17:10:02 UTC

Augmented Inverse Probability Weighting (AIPW)

Description

An R6Class of AIPW for estimating the average causal effects with users' inputs of exposure, outcome, covariates and related libraries for estimating the efficient influence function.

Details

An AIPW object is constructed by new() with users' inputs of data and causal structures, then it fit() the data using the libraries in Q.SL.library and g.SL.library with k_split cross-fitting, and provides results via the summary() method. After using fit() and/or summary() methods, propensity scores and inverse probability weights by exposure status can be examined with plot.p_score() and plot.ip_weights(), respectively.

If outcome is missing, analysis assumes missing at random (MAR) by estimating propensity scores of I(A=a, observed=1) with all covariates W. (W.Q and W.g are disabled.) Missing exposure is not supported.

See examples for illustration.

Value

AIPW object

Constructor

AIPW$new(Y = NULL, A = NULL, W = NULL, W.Q = NULL, W.g = NULL, Q.SL.library = NULL, g.SL.library = NULL, k_split = 10, verbose = TRUE, save.sl.fit = FALSE)

Constructor Arguments

Constructor Argument Details

W, W.Q & W.g

It can be a vector, matrix or data.frame. If and only if W == NULL, W would be replaced by W.Q and W.g.

Q.SL.library & g.SL.library

Machine learning algorithms from SuperLearner libraries or sl3 learner object (Lrnr_base)

k_split

It ranges from 1 to number of observation-1. If k_split=1, no cross-fitting; if k_split>=2, cross-fitting is used (e.g., k_split=10, use 9/10 of the data to estimate and the remaining 1/10 leftover to predict). NOTE: it's recommended to use cross-fitting.

save.sl.fit

This option allows users to save the fitted sl object (libs$Q.fit & libs$g.fit) for debug use. Warning: Saving the SuperLearner fitted object may cause a substantive storage/memory use.

Public Methods

Public Variables

Public Variable Details

stratified_fit

An indicator for whether the outcome model is fitted stratified by exposure status in the fit() method. Only when using stratified_fit() to turn on stratified_fit = TRUE, summary outputs average treatment effects among the treated and the controls.

obs_est

After using fit() and summary() methods, this list contains the propensity scores (p_score), counterfactual predictions (mu, mu1 & mu0) and efficient influence functions (aipw_eif1 & aipw_eif0) for later average treatment effect calculations.

g.plot

This plot is generated by ggplot2::geom_density

ip_weights.plot

This plot uses truncated propensity scores stratified by exposure status (ggplot2::geom_boxplot)

References

Zhong Y, Kennedy EH, Bodnar LM, Naimi AI (2021). AIPW: An R Package for Augmented Inverse Probability Weighted Estimation of Average Causal Effects. American Journal of Epidemiology.

Robins JM, Rotnitzky A (1995). Semiparametric efficiency in multivariate regression models with missing data. Journal of the American Statistical Association.

Chernozhukov V, Chetverikov V, Demirer M, et al (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal.

Kennedy EH, Sjolander A, Small DS (2015). Semiparametric causal inference in matched cohort studies. Biometrika.

Examples

library(SuperLearner)
library(ggplot2)

#create an object
aipw_sl <- AIPW$new(Y=rbinom(100,1,0.5), A=rbinom(100,1,0.5),
                    W.Q=rbinom(100,1,0.5), W.g=rbinom(100,1,0.5),
                    Q.SL.library="SL.mean",g.SL.library="SL.mean",
                    k_split=1,verbose=FALSE)

#fit the object
aipw_sl$fit()
# or use `aipw_sl$stratified_fit()` to estimate ATE and ATT/ATC

#calculate the results
aipw_sl$summary(g.bound = 0.025)

#check the propensity scores by exposure status after truncation
aipw_sl$plot.p_score()

Augmented Inverse Probability Weighting Base Class (AIPW_base)

Description

A base class for AIPW that implements the common methods, such as summary() and plot.p_score(), inheritted by AIPW and AIPW_tmle class

Format

R6 object.

Value

AIPW base object

See Also

AIPW and AIPW_tmle

Augmented Inverse Probability Weighting (AIPW) uses tmle or tmle3 as inputs

Description

AIPW_nuis class for users to manually input nuisance functions (estimates from the exposure and the outcome models)

Details

Create an AIPW_nuis object that uses users' input nuisance functions from the exposure model P(A| W), and the outcome models P(Y| do(A=0), W) and P(Y| do(A=1), W.Q):

\psi(a) = E{[ I(A=a) / P(A=a|W) ] * [Y-P(Y=1|A,W)] + P(Y=1| do(A=a),W) }

Note: If outcome is missing, replace (A=a) with (A=a, observed=1) when estimating the propensity scores.

Value

AIPW_nuis object

Constructor

AIPW$new(Y = NULL, A = NULL, tmle_fit = NULL, verbose = TRUE)

Constructor Arguments

Public Methods

Public Variables

Public Variable Details

stratified_fit

An indicator for whether the outcome model is fitted stratified by exposure status in thefit() method. Only when using stratified_fit() to turn on stratified_fit = TRUE, summary outputs average treatment effects among the treated and the controls.

obs_est

This list includes propensity scores (p_score), counterfactual predictions (mu, mu1 & mu0) and efficient influence functions (aipw_eif1 & aipw_eif0)

g.plot

This plot is generated by ggplot2::geom_density

ip_weights.plot

This plot uses truncated propensity scores stratified by exposure status (ggplot2::geom_boxplot)

Augmented Inverse Probability Weighting (AIPW) uses tmle or tmle3 as inputs

Description

AIPW_tmle class uses a fitted tmle or tmle3 object as input

Details

Create an AIPW_tmle object that uses the estimated efficient influence function from a fitted tmle or tmle3 object

Value

AIPW_tmle object

Constructor

AIPW$new(Y = NULL, A = NULL, tmle_fit = NULL, verbose = TRUE)

Constructor Arguments

Public Methods

Public Variables

Public Variable Details

obs_est

This list extracts from the fitted tmle or tmle3 object. It includes propensity scores (p_score), counterfactual predictions (mu, mu1 & mu0) and efficient influence functions (aipw_eif1 & aipw_eif0)

g.plot

This plot is generated by ggplot2::geom_density

ip_weights.plot

This plot uses truncated propensity scores stratified by exposure status (ggplot2::geom_boxplot)

Examples

## Not run: 
vec <- function() sample(0:1,100,replace = TRUE)
df <- data.frame(replicate(4,vec()))
names(df) <- c("A","Y","W1","W2")

## From tmle
library(tmle)
library(SuperLearner)
tmle_fit <- tmle(Y=df$Y,A=df$A,W=subset(df,select=c("W1","W2")),
                 Q.SL.library="SL.glm",
                 g.SL.library="SL.glm",
                 family="binomial")
AIPW_tmle$new(A=df$A,Y=df$Y,tmle_fit = tmle_fit,verbose = TRUE)$summary()


## From tmle3
# tmle3 simple implementation
library(tmle3)
library(sl3)
node_list <- list(A = "A",Y = "Y",W = c("W1","W2"))
or_spec <- tmle_OR(baseline_level = "0",contrast_level = "1")
tmle_task <- or_spec$make_tmle_task(df,node_list)
lrnr_glm <- make_learner(Lrnr_glm)
sl <- Lrnr_sl$new(learners = list(lrnr_glm))
learner_list <- list(A = sl, Y = sl)
tmle3_fit <- tmle3(or_spec, data=df, node_list, learner_list)

# parse tmle3_fit into AIPW_tmle class
AIPW_tmle$new(A=df$A,Y=df$Y,tmle_fit = tmle3_fit,verbose = TRUE)$summary()

## End(Not run)

Repeated Crossfitting Procedure for AIPW

Description

An R6Class that allows repeated crossfitting procedure for an AIPW object

Details

See examples for illustration.

Value

AIPW object

Constructor

Repeated$new(aipw_obj = NULL)

Constructor Arguments

Public Methods

Public Variables

Public Variable Details

repeated_estimates

Estimates from num_reps cross-fitting.

result

Summarised estimates from “repeated_estimates' using median methods.

References

Zhong Y, Kennedy EH, Bodnar LM, Naimi AI (2021). AIPW: An R Package for Augmented Inverse Probability Weighted Estimation of Average Causal Effects. American Journal of Epidemiology.

Robins JM, Rotnitzky A (1995). Semiparametric efficiency in multivariate regression models with missing data. Journal of the American Statistical Association.

Chernozhukov V, Chetverikov V, Demirer M, et al (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal.

Kennedy EH, Sjolander A, Small DS (2015). Semiparametric causal inference in matched cohort studies. Biometrika.

Examples

library(SuperLearner)
library(ggplot2)

#create an object
aipw_sl <- AIPW$new(Y=rbinom(100,1,0.5), A=rbinom(100,1,0.5),
                    W.Q=rbinom(100,1,0.5), W.g=rbinom(100,1,0.5),
                    Q.SL.library="SL.mean",g.SL.library="SL.mean",
                    k_split=2,verbose=FALSE)

#create a repeated crossfitting object from the previous step
repeated_aipw_sl <- Repeated$new(aipw_sl)

#fit repetitively (stratified = TRUE will use stratified_fit() method in AIPW class)
repeated_aipw_sl$repfit(num_reps = 3, stratified = FALSE)

#summarise the results
repeated_aipw_sl$summary_median()

AIPW wrapper function

Description

A wrapper function for AIPW$new()$fit()$summary()

Usage

aipw_wrapper(
  Y,
  A,
  verbose = TRUE,
  W = NULL,
  W.Q = NULL,
  W.g = NULL,
  Q.SL.library,
  g.SL.library,
  k_split = 10,
  g.bound = 0.025,
  stratified_fit = FALSE
)

Arguments

Value

A fitted AIPW object with summarised results

See Also

AIPW

Examples

library(SuperLearner)
aipw_sl <- aipw_wrapper(Y=rbinom(100,1,0.5), A=rbinom(100,1,0.5),
                    W.Q=rbinom(100,1,0.5), W.g=rbinom(100,1,0.5),
                    Q.SL.library="SL.mean",g.SL.library="SL.mean",
                    k_split=1,verbose=FALSE)

Simulated Observational Study

Description

Datasets were simulated using baseline covariates (sampling with replacement) from the Effects of Aspirin in Gestation and Reproduction (EAGeR) study. Data generating mechanisms were described in our manuscript (Zhong et al. (inpreparation), Am. J. Epidemiol.). True marginal causal effects on risk difference, log risk ratio and log odds ratio scales were attached to the dataset attributes (true_rd, true_logrr,true_logor).

Usage

data(eager_sim_obs)

Format

An object of class data.frame with 200 rows and 8 columns:

sim_Y

binary, simulated outcome which is condition on all other covariates in the dataset

sim_A

binary, simulated exposure which is conditon on all other covarites expect sim_Y.

eligibility

binary, indicator of the eligibility stratum

loss_num

count, number of prior pregnancy losses

age

continuous, age in years

time_try_pregnant

count, months of conception attempts prior to randomization

BMI

continuous, body mass index

meanAP

continuous, mean arterial blood pressure

References

Schisterman, E.F., Silver, R.M., Lesher, L.L., Faraggi, D., Wactawski-Wende, J., Townsend, J.M., Lynch, A.M., Perkins, N.J., Mumford, S.L. and Galai, N., 2014. Preconception low-dose aspirin and pregnancy outcomes: results from the EAGeR randomised trial. The Lancet, 384(9937), pp.29-36.

Zhong, Y., Naimi, A.I., Kennedy, E.H., (In preparation). AIPW: An R package for Augmented Inverse Probability Weighted Estimation of Average Causal Effects. American Journal of Epidemiology

See Also

eager_sim_rct

Simulated Randomized Trial

Description

Datasets were simulated using baseline covariates (sampling with replacement) from the Effects of Aspirin in Gestation and Reproduction (EAGeR) study.

Usage

data(eager_sim_rct)

Format

An object of class data.frame with 1228 rows and 8 columns:

sim_Y

binary, simulated outcome which is condition on all other covariates in the dataset

sim_T

binary, simulated treatment which is condition on eligibility only.

eligibility

binary, indicator of the eligibility stratum

loss_num

count, number of prior pregnancy losses

age

continuous, age in years

time_try_pregnant

count, months of conception attempts prior to randomization

BMI

continuous, body mass index

meanAP

continuous, mean arterial blood pressure

References

Schisterman, E.F., Silver, R.M., Lesher, L.L., Faraggi, D., Wactawski-Wende, J., Townsend, J.M., Lynch, A.M., Perkins, N.J., Mumford, S.L. and Galai, N., 2014. Preconception low-dose aspirin and pregnancy outcomes: results from the EAGeR randomised trial. The Lancet, 384(9937), pp.29-36.

Zhong, Y., Naimi, A.I., Kennedy, E.H., (In preparation). AIPW: An R package for Augmented Inverse Probability Weighted Estimation of Average Causal Effects. American Journal of Epidemiology

See Also

eager_sim_obs

Fit the data to the AIPW object

Description

Fitting the data into the AIPW object with/without cross-fitting to estimate the efficient influence functions

Value

A fitted AIPW object with obs_est and libs (public variables)

R6 Usage

$fit()

See Also

AIPW

Plot the inverse probability weights using truncated propensity scores by exposure status

Description

Plot and check the balance of propensity scores by exposure status

Value

ip_weights.plot (public variable): A box plot of inverse probability weights using truncated propensity scores by exposure status (ggplot2::geom_boxplot)

R6 Usage

$plot.ip_weights()

See Also

AIPW and AIPW_tmle

Plot the propensity scores by exposure status

Description

Plot and check the balance of propensity scores by exposure status

Value

g.plot (public variable): A density plot of propensity scores by exposure status (ggplot2::geom_density)

R6 Usage

$plot.p_plot()

See Also

AIPW and AIPW_tmle

Fit the data to the AIPW object repeatedly

Description

Fitting the data into the AIPW object with cross-fitting repeatedly to obtain multiple estimates from repetitions to avoid randomness due to splits in cross-fitting

Arguments

Value

A Repeated object with repeated_estimates (estimates from num_reps times repetition)

R6 Usage

$repfit(num_reps = 20, stratified = FALSE)

References

Chernozhukov V, Chetverikov V, Demirer M, et al (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal.

See Also

Repeated and AIPW

Fit the data to the AIPW object stratified by A for the outcome model

Description

Fitting the data into the AIPW object with/without cross-fitting to estimate the efficient influence functions. Outcome model is fitted, stratified by exposure status A

Value

A fitted AIPW object with obs_est and libs (public variables)

R6 Usage

$stratified_fit.AIPW()

See Also

AIPW

Summary of the average treatment effects from AIPW

Description

Calculate average causal effects in RD, RR and OR in the fitted AIPW or AIPW_tmle object using the estimated efficient influence functions

Arguments

Value

estimates and result (public variables): Risks, Average treatment effect in RD, RR and OR.

R6 Usage

$summary(g.bound = 0.025)
$summary(g.bound = c(0.025,0.975))

See Also

AIPW and AIPW_tmle

Summary of the repeated_estimates from repfit() in the Repeated object using median methods.

Description

From repeated_estimates, calculate the median estimate (median(Estimates)), median SE (median(SE)), SE adjusting for variations across num_reps times, and 95% CI using SE adjusting for SE adjusted for variability.

Value

repeated_results and result (public variables).

R6 Usage

$summary_median.Repeated()

References

Chernozhukov V, Chetverikov V, Demirer M, et al (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal.

See Also

Repeated and AIPW