flintyR - Flexible and Interpretable Non-Parametric Tests of Exchangeability

Given a multivariate dataset and some knowledge about the dependencies between its features, it is important to ensure the observations or individuals are exchangeable before fitting a model to the data in order to make inferences from it, or assigning randomized treatments in order to estimate treatment effects. This package provides a flexible non-parametric test of exchangeability, allowing the user to specify the feature dependencies by hand. It can be used directly to evaluate whether a sample is exchangeable, and can also be piped into larger procedures that require exchangeable samples as outputs (e.g., clustering or community detection). See Aw, Spence and Song (2021+) for the accompanying paper.