Time for a change: a tutorial for comparing multiple classifiers through\n Bayesian analysis

Abstract

The machine learning community adopted the use of null hypothesis\nsignificance testing (NHST) in order to ensure the statistical validity of\nresults. Many scientific fields however realized the shortcomings of\nfrequentist reasoning and in the most radical cases even banned its use in\npublications. We should do the same: just as we have embraced the Bayesian\nparadigm in the development of new machine learning methods, so we should also\nuse it in the analysis of our own results. We argue for abandonment of NHST by\nexposing its fallacies and, more importantly, offer better - more sound and\nuseful - alternatives for it.\n