A Kruskal-Wallis test is added to ion

I’ve added a script to ion to perform a Kruskal-Wallis test for each row in a data matrix (ion$kruskal_test_3g). This statistical test is a non-parametric test and can be seen as a multi-group generalization of the Mann-Whitney test (or Wilcoxon test) for two-group comparisons.

A non-parametric test is robust and applicable to various types of data including spectral count data and intensity-based quantification in proteomics. Nevertheless, it requires a larger number of samples than a parametric test such as the beta-binomial test or the t-test. I expect that 10 samples per group are sufficient. I’ve read somewhere that 15 should be the minimum. But I have also seen studies using a non-parametric test for 5 samples per group. In short, there is no hard rule that I am aware of.