Is it time to de-normalize the normalization procedures based on ratio variables?

Abstract

Ratio variables are created by dividing one variable by another variable. This procedure is usually done with the aim of “normalizing” for the denominator variable, to obtain associations adjusted for that confounding variable, or to calculate ranges of normative values. Despite its widespread use, ratio variables bring with them many statistical problems. In this editorial, the reader will be introduced to the statistical assumptions of using ratio variables, the problems that arise when those assumptions are not met, and how to use proper regression models to achieve the desired objectives without the need to use ratio variables.