In youmisuk/CURobustML: Robust Machine Learning Under Cluster-Level Unmeasured Confounding

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

library(h2o)
localH2O = h2o.init()
h2o.removeAll() # remove all objects on the H2O Cluster
h2o.no_progress() # without progress bar

CURobustML

A family of machine learning (ML) estimators for handling unmeasured cluster-level confounders in R. It provides a general approach to estimate causal effects in the presence of cluster-level unmeasured confounders in multilevel observational data. In particular, we leverage modern ML methods and exploit a fundamental nature regarding cluster-level unmeasured confounders to estimate the conditional average treatment effect (CATE) and the average treatment effect (ATE). See Suk and Kang (2020) <doi:10.31234/osf.io/t7vbz> for details.

You can also evaluate the proposed estimators in Shiny app, https://youmi.shinyapps.io/curobustml.

Installation

You can install the most recent version of CURobustML from GitHub with:

install.packages("devtools")
library(devtools)

install_github("youmisuk/CURobustML")

You should also start and connect to H2O instance via package h2o to implement CURobustML.

library(h2o)
localH2O = h2o.init()
h2o.removeAll() # clean slate - just in case the cluster was already running

Or, use the setup function.

library(CURobustML)

CURobustML.setup()

Example

• The doubly robust estimator with proxy regression with two-level data

DRPRcomb.rslt <- DRPRcomb(Y=twolevel_data$Y, Z=twolevel_data$Z, interZ=(~ W1),
                      X=twolevel_data[, c("X1", "X2", "X3", "W1")], ID=twolevel_data$id, data=twolevel_data)

summary(DRPRcomb.rslt)

# with final predictions
DR(Y=twolevel_data$Y, Z=twolevel_data$Z, interZ=(~ W1), Z.hat=DRPRcomb.rslt$Z.hat,
   Y1.hat=DRPRcomb.rslt$Y1.hat, Y0.hat=DRPRcomb.rslt$Y0.hat, data=twolevel_data)

# with predictions from glm wtih fixed effects of clusters
DR(Y=twolevel_data$Y, Z=twolevel_data$Z, interZ=(~ W1), Z.hat=DRPRcomb.rslt$Z.hats$Ztest.hat_glm_1,
   Y1.hat=DRPRcomb.rslt$Y1.hats$Y1test.hat_glm_1, Y0.hat=DRPRcomb.rslt$Y0.hats$Y0test.hat_glm_1, data=twolevel_data)

• The double demeaning estimator with cross-classified data

DDcomb.rslt <- DDcomb(Y=crossclassified_data$Y, Z=crossclassified_data$Z, interZ=(~ W1),
                      X=crossclassified_data[, c("X1", "X2", "X3", "W1", "Q1")],
                      ID=crossclassified_data$f12id, data=crossclassified_data)

summary(DDcomb.rslt)

# with final predictions
DD(Y=crossclassified_data$Y, Z=crossclassified_data$Z, interZ=(~ W1), ID=crossclassified_data$f12id,
   Z.hat=DDcomb.rslt$Z.hat, Y0.hat=DDcomb.rslt$Y0.hat, data=crossclassified_data)

# with predictions from glm
DD(Y=crossclassified_data$Y, Z=crossclassified_data$Z, interZ=(~ W1), ID=crossclassified_data$f12id,
   Z.hat=DDcomb.rslt$Z.hats$Ztest.hat_glm, Y0.hat=DDcomb.rslt$Y0.hats$Y0test.hat_glm, data=crossclassified_data)

# with predictions from deep learning
DD(Y=crossclassified_data$Y, Z=crossclassified_data$Z, interZ=(~ W1), ID=crossclassified_data$f12id,
   Z.hat=DDcomb.rslt$Z.hats$Ztest.hat_dl, Y0.hat=DDcomb.rslt$Y0.hats$Y0test.hat_dl, data=crossclassified_data)

youmisuk/CURobustML documentation built on Sept. 11, 2022, 11:04 a.m.