A Practical Guide to Bayesian Statistics in Laboratory Medicine.

Abstract

Statistical analyses form a fundamental part of causal inference in the experimental sciences. The statistical paradigm most commonly taught to science students around the world is that of frequentism, with a particular emphasis on the null hypothesis significance testing borne by the work of Neyman and Pearson in the early 20th century. This paradigm is often lauded as being the most objective of methods and remains commonplace in scientific journals. Despite its widespread use-and, indeed, requirement for publication in some journals-this paradigm has received substantial criticism in recent decades, and its impact on scientific publishing has been subjected to deeper scrutiny in response to the replication crisis in the psychological and medical sciences. It has been posited that the increasing use of the Bayesian statistical paradigm, made more accessible through technological advances in the last few decades, may have an important role to play in rendering research and statistical inference more robust, transparent, and reproducible. These methods can have a steep learning curve, and thus this paper seeks to introduce those working within clinical laboratories to the Bayesian paradigm of statistical analysis and provides worked examples of the Bayesian analysis of data commonly encountered in laboratory medicine using freely available, open source tools.