Wilcoxon Rank-Based Tests for Clustered Data with R Package clusrank

TL;DR

The paper introduces the R package clusrank, which implements various Wilcoxon rank-based tests for clustered data, providing a unified, flexible, and user-friendly tool that includes permutation-based exact tests and supports comparison of methods.

Contribution

The package consolidates recent methodological developments for rank-based tests on clustered data into a single, accessible R package with a unified interface.

Findings

The package enables flexible application of rank-based tests to clustered data.

It includes permutation-based exact tests for certain methods.

Illustrations demonstrate practical use and comparison of methods.

Abstract

Wilcoxon Rank-based tests are distribution-free alternatives to the popular two-sample and paired t-tests. For independent data, they are available in several R packages such as stats and coin. For clustered data, in spite of the recent methodological developments, there did not exist an R package that makes them available at one place. We present a package clusrank where the latest developments are implemented and wrapped under a unified user-friendly interface. With different methods dispatched based on the inputs, this package offers great flexibility in rank-based tests for various clustered data. Exact tests based on permutations are also provided for some methods. Details of the major schools of different methods are briefly reviewed. Usages of the package clusrank are illustrated with simulated data as well as a real dataset from an ophthalmological study. The package also enables convenient comparison between selected methods under settings that have not been studied before and the results are discussed.