Paul R. Yarnold
Optimal Data Analysis, LLC
Compared to the Mann-Whitney U test, UniODA explicitly maximizes model accuracy for a specific sample and hypothesis; identifies the optimal threshold discriminating the groups; yields invariant findings over monotonic transformations of the data; indexes model effect strength on an absolute scale ranging from 0 (accuracy expected for the sample and hypothesis by chance) to 100 (perfect accuracy); and estimates cross-generalizability of the findings if the model is applied to classify an independent random sample.