Bayesian nonparametric testing - is it also more robust to outliers?
Hello,
In the classical NHST, nonparametric tests like Kendall's or Spearman's rank correlations or Mann–Whitney U test are not only used to non-normally distributed data but also to sets that seems normally distributed but with present of "outliers", in order to receive more robustness to them as compared to the parametric counterparts.
I've started to wonder if Bayesian versions of nonparametric tests (in example implemented in JASP Bayesian Kendall's tau-b or Bayesian Mann–Whitney) have the same, "robustness to outliers" properties?
Thank you in advance for any guidance.
Comments
Yes they do. Ranks are much more robust in general.