Bayesian and frequentist testing for differences between two groups with parametric and nonparametric two‐sample tests. (13th July 2020)

• Record Type: Journal Article
• Authors: Kelter, Riko
• Is Part Of: Wiley interdisciplinary reviews. Volume 13:Number 6(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-07-13
• Subjects: Bayesian two‐sample tests -- Mann–Whitney U test -- null hypothesis significance testing -- Student's t‐test -- testing for differences between two groups; Mathematical statistics -- Data processing -- Periodicals; Science -- Data processing -- Periodicals; Social sciences -- Data processing -- Periodicals; Mathematical statistics -- Periodicals; 519.50285
• DOI: 10.1002/wics.1523
• Languages: English
• ISSNs: 1939-5108

Abstract

Abstract: Testing for differences between two groups is one of the scenarios most often faced by scientists across all domains and is particularly important in the medical sciences and psychology. The traditional solution to this problem is rooted inside the Neyman–Pearson theory of null hypothesis significance testing and uniformly most powerful tests. In the last decade, a lot of progress has been made in developing Bayesian versions of the most common parametric and nonparametric two‐sample tests, including Student's t ‐test and the Mann–Whitney U test. In this article, we review the underlying assumptions, models and implications for research practice of these Bayesian two‐sample tests and contrast them with the existing frequentist solutions. Also, we show that in general Bayesian and frequentist two‐sample tests have different behavior regarding the type I and II error control, which needs to be carefully balanced in practical research. This article is categorized under: Statistical and Graphical Methods of Data Analysis > Bayesian Methods and Theory Statistical and Graphical Methods of Data Analysis > Monte Carlo Methods Statistical and Graphical Methods of Data Analysis > Markov Chain Monte Carlo Abstract : Visual representation of a Bayesian two‐sample t ‐test. Bayesian two‐sample tests have recently becomes increasingly popular through the on‐going discussion about statistical significance, and we review and contrast them with existing frequentist tests, also investigating their type I and II error control.