Hi,
thanks for the great waddR package!
I am currently working on a paper on improving the accuracy of small permutation p-values. waddR is one of the tools we use to demonstrate our workflow, which requires access to both the observed and the permuted test statistics. At the moment, I rely on a small custom implementation to retain the permutation test statistics, since the current version of waddR returns only the observed, but not the permuted, Wasserstein distances.
For reproducibility and transparency, it would be much nicer if waddR directly returned the permutation distributions. My current workaround is to store them as an attribute (e.g., attr(RES, "wass.values") <- wass.vals.mat) to not change the output itself, but there might be cleaner solutions.
Thanks for considering this enhancement!
Best,
Stefanie
Hi,
thanks for the great waddR package!
I am currently working on a paper on improving the accuracy of small permutation p-values. waddR is one of the tools we use to demonstrate our workflow, which requires access to both the observed and the permuted test statistics. At the moment, I rely on a small custom implementation to retain the permutation test statistics, since the current version of waddR returns only the observed, but not the permuted, Wasserstein distances.
For reproducibility and transparency, it would be much nicer if waddR directly returned the permutation distributions. My current workaround is to store them as an attribute (e.g.,
attr(RES, "wass.values") <- wass.vals.mat) to not change the output itself, but there might be cleaner solutions.Thanks for considering this enhancement!
Best,
Stefanie