mlbalance

Balance Permutation Test (Rametta, Fuller 2024)

https://github.com/cetialphafive/mlbalance

Keywords

Repository

Balance Permutation Test (Rametta, Fuller 2024)

README.Rmd

---
output: github_document
---

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

# MLbalance (alpha version) 

MLbalance implements a novel machine learning balance test, the balance permutation test, for experiments with binary, multiarm, and continuous treatments. The purpose of this test is to detect failures of random assignment or imbalance across treatment arms. For more detail, see Rametta and Fuller (2024). 

This package is in alpha, any recommendations or comments welcome in the issues section.  

## Installation

Stable version is be available on Github here: 

``` r 
# install.packages("pak")
pak::pak("samjfuller/MLbalance")
```

You can install the development version of MLbalance from [GitHub](https://github.com/CetiAlphaFive/MLbalance) with:

``` r
# install.packages("pak")
pak::pak("CetiAlphaFive/MLbalance")
```

## Binary Treatment Example 

Here is a basic example demonstrating the balance permutation test on a simulated binary treatment DGP with multidimensional contamation of the treatment assignment.

```{r example}
# install.packages("randomizr")
library(MLbalance)
#
set.seed(1995)
#
# Simple simulation 
n <- 1000
p <- 20
X <- matrix(rnorm(n*p,0,1),n,p)
w_real <- rbinom(n, 1, ifelse(.021 + abs(.4*X[,4] - .5*X[,8]) < 1, .021 + abs(.4*X[,4] - .5*X[,8]), 1))
# install.packages("randomizr")
w_sim <- randomizr::complete_ra(N = n,m = sum(w_real))
e <- rnorm(n,0,1)
y <- 2*w_real*X[,4] + 3*X[,2] -2*X[,8] + e
df <- data.frame(y,w_real,w_sim,X)
#
r.check <- random_check(W_real = df$w_real, #real treatment assignment vector 
                        W_sim  = df$w_sim, #simulated vector, comment out this argument to use permutated real assignment vector instead 
                        X      = subset(df,select = -c(y,w_real,w_sim)) #matrix of pretreatment covariates (or any covariates that SHOULD NOT be related to the assignment process/mechanism
             ); r.check$plot
# to see variables important for predicting assignment, check r.check$imp.predictors 
```

Owner

Citation (CITATION.cff)

cff-version: 1.2.0
message: "If you use this software, please cite it as below."
authors:
- family-names: "Rametta"
  given-names: "Jack"
  orcid: "https://orcid.org/0000-0002-9841-146X"
- family-names: "Fuller"
  given-names: "Sam"
  orcid: "https://orcid.org/0000-0002-9670-1092"
title: "MLbalance"
version: 0.1
date-released: 2024-08-02
url: "https://github.com/samjfuller/MLbalance"

GitHub Events