R-package

Installation

To use iArt in R, install it using the devtools package:

# Install iArt from GitHub
devtools::install_github("Imputation-Assisted-Randomization-Tests/iArt")

Conducting Randomization Tests with iArt

The iArt.test function in the iArt package allows for conducting finite-population-exact randomization tests in design-based causal studies with missing outcomes.

Basic Usage Example

To perform a basic randomization test in R:

library(iArt)

Z <- c(1, 1, 1, 1, 0, 0, 0, 0)

X <- matrix(c(5.1, 3.5, 4.9, NA, 4.7, 3.2, 4.5, NA, 7.2, 2.3, 8.6, 3.1, 6.0, 3.6, 8.4, 3.9), ncol = 2)

Y <- matrix(c(4.4, 0.5, 4.3, 0.7, 4.1, NA, 5.0, 0.4, 1.7, 0.1, NA, 0.2, 1.4, NA, 1.7, 0.4), ncol = 2)

result <- iArt.test(Z, X, Y, L = 1000, verbose = TRUE)

print(result)

Different Imputation Methods in iArt

The iArt package supports various methods for imputing missing data in R. Each method can be specified using the G parameter in the iArt.test function. It is slightly different from the Python version, because the R version uses the missforest and mice, whereas the Python version uses the IterativeImputer

To specify an imputation method in R, set the G parameter in the iArt.test function to one of the options.

MICE (`’mice’`):
Multiple Imputation by Chained Equations (MICE) is a popular imputation method in R. It is a fully conditional specification method that creates multiple imputations by filling in missing data multiple times using a specified imputation model.
result <- iArt.test(Z, X, Y, G = 'mice', L = 1000, verbose = TRUE)
You can find more about the mice package on its CRAN page here: mice.
MissForest (`’missforest’`):
MissForest is a non-parametric imputation method in R. It is a non-parametric imputation method that uses a random forest to impute missing data.
result <- iArt.test(Z, X, Y, G = 'missforest', L = 1000, verbose = TRUE)
For more information on missforest, visit the CRAN page: missforest.

Handling Strata in the Data

Incorporate strata into your analysis in R using the S parameter in the iArt.test function. All strata should be in ascending order, and the stratum index for each data point should be specified in the same order as the data points

# S vector indicates the stratum index for each data point
S <- c(0, 0, 1, 1, 1, 2, 2, 2)

# Incorporating strata into the analysis
result <- iArt.test(Z, X, Y, S = S, L = 1000, verbose = TRUE)

Covariate Adjustment

Enable covariate adjustment in R by setting the covariate_adjustment parameter to TRUE in the iArt.test function. This will use linear regression to adjust for covariates in the analysis.

# Conducting a randomization test with covariate adjustment
result <- iArt.test(Z, X, Y, covariate_adjustment = TRUE, L = 1000, verbose = TRUE)

Specifying an Alternative Hypothesis:

To specify an alternative hypothesis in R, set the alternative parameter in the iArt.test function to either ‘greater’, ‘less’, or ‘two-sided’.

# For a one-sided test greater than
result <- iArt.test(Z, X, Y, alternative = "greater", L = 1000, verbose = TRUE)

# For a one-sided test less than
result <- iArt.test(Z, X, Y, alternative = "less", L = 1000, verbose = TRUE)

# For a two-sided test
result <- iArt.test(Z, X, Y, alternative = "two-sided", L = 1000, verbose = TRUE)

Setting a Random State for Reproducibility:

Set a random state for reproducibility in R:

# Setting a seed for reproducibility
set.seed(42)
result <- iArt.test(Z, X, Y, L = 1000, verbose = TRUE)