Statistical approaches to make sense of data in biology and medicine

Abstract

There are four major paradigms in statistics: Frequentist, Bayesian, likelihood, and modeling. A quadrangle approach that makes use of all these four paradigms is proposed to get a complete understanding of any biological phenomenon. Each of these paradigms can be used to study different aspects of a biological phenomenon. The elements are defined here as an observer, observed, and context, and the model generated should have information derived from these three elements. They can be analyzed, respectively, by Bayesian, frequentist, likelihood, and modeling methods. There is a continuous debate on frequentist and Bayesian approaches in statistics. Biologists often use frequentist methods whereas clinicians are interested in Bayesian methods. In this article, the debate on both these approaches has been discussed in light of understanding uncertainty. The Dempster-Shafer theory addresses the relationship between belief and plausibility but has been criticized for producing counterintuitive results in conflict situations. It is argued here that this can be resolved by inferring that frequentist and Bayesian approaches are inverse to each other.