Last update:


Arch Hellen Med, 34(5), September-October 2017, 622-627


Bayes factor: A brief review of an alternative way of making statistical decisions

D. Panaretos, D.B. Panagiotakos
Department of Nutrition and Dietetics, School of Health Science and Education, Harokopio University, Athens, Greece

This review briefly introduces the two schools of statistical inference: Bayesian and Frequentist. Through arguable evidence, the conclusion is reached that Bayesian methodology can contribute decisively, leading to valid inferences in an inductive way. This is because the Bayesian methodology can combine prior (by the prior probability of the null hypothesis to be true) with current evidence (by the calculation of the Bayes factor) to calculate the definitive probability of the null hypothesis to be true. Hypothesis testing and p-values are often used wrongly by researchers for deriving conclusions. In contrast to the Bayes theorem, hypothesis tests express probability as the relative frequency of the facts that are described in the statistical hypothesis and therefore cannot be used to draw conclusions based on the data of a single study, although it can provide important information after the repetition of the same study in the long term. It is suggested that there should be correct usage of the p-value, in combination with the Bayes factor, in order to derive credible inferences.

Key words: Bayes factor, Hypothesis testing, Likelihood, p-value, Strength of association.

© Archives of Hellenic Medicine