The new frontier of statistics: Modern machine learning approaches as alternatives to traditional statistical tests in biological, clinical, and epidemiological research with a focus on cardiac event prediction

Abstract

As the complexity and volume of biological and clinical data increase, traditional statistical methods, such as logistic regression, discriminant analysis, analysis of variance (ANOVA), and multivariate analysis, often fall short of capturing the intricate patterns needed for accurate prediction and classification. Here, we explore alternative analytical frameworks rooted in modern machine learning (ML) techniques that offer enhanced capabilities for diverse biomedical applications. For example, these frameworks demonstrate superior predictive performance for cardiac events compared with classical logistic regression. However, challenges, interpretability, and future directions are important considerations when facing this new frontier. Moreover, systematically integrating these advanced computational tools into routine clinical and epidemiological research is imperative. This co-authored column forms part of the “Statistics Series” and builds on A simple guide to analyse data by Prof. Libhaber.⁽¹⁾