The robusTest package: two-sample tests revisited

TL;DR

The robusTest R package provides robust, corrected versions of common bivariate statistical tests, addressing limitations of classical methods and demonstrating improved performance through simulations and real data application.

Contribution

This paper introduces robust alternatives to classical bivariate tests in the robusTest package, improving robustness and calibration.

Findings

Robust tests outperform classical versions in simulations.

Robust correlation tests show better calibration.

Application on real data illustrates practical advantages.

Abstract

The R package robusTest offers corrected versions of several common tests in bivariate statistics. We point out the limitations of these tests in their classical versions, some of which are well known such as robustness or calibration problems, and provide simple alternatives that can be easily used instead. The classical tests and theirs robust alternatives are compared through a small simulation study. The latter emphasizes the superiority of robust versions of the test of interest. Finally, an illustration of correlation's tests on a real data set is also provided.