Control of third variables: mediation, interaction and confusion

Abstract

In health sciences, bio-psycho-social and cultural phenomena cannot be explained by a relationship between two variables or a group of predictors and an outcome. Relationships between variables form complex causal networks where multiple factors interact simultaneously and multidimensionally. In this scenario, it is essential to integrate these interrelationships into the design and statistical analysis by examining the processes of mediation, interaction, and confusion. This article aims to serve as a guide for health professionals who wish to begin analysing these processes with a twofold purpose: on the one hand, to provide consumers of scientific information with the necessary basis for their correct understanding and interpretation; on the other, to provide researchers with the methodological tools to analyze, describe, and control these processes, allowing for more robust and unbiased estimates.