datathin R package splits a random variable (or a vector or matrix of random variables) into an independent training set and a test set using the methodology introduced in Neufeld et al., 2023 (link to preprint).
Make sure that
remotes is installed by running
install.packages("remotes"), then type
For now, you can check our our introductory tutorial, which gives basic examples of how to apply data thinning under a variety of distributional assumptions. You can also check out our unsupervised learning tutorial, which shows how data thinning can be applied to estimate the number of clusters in normally distributed data.
More tutorials for this package are coming soon! We will provide examples of how to use
datathin for tasks such as model evaluation and inference after model selection under a variety of distributional assumptions.
To learn more, check out our preprint.
To reproduce the figures from our preprint, please see the following repository: https://github.com/anna-neufeld/datathin_paper.