A conceptual and empirical examination of justifications for dichotomization.

TLDR

Researchers frequently dichotomize continuous measures despite widespread criticism of the practice. The study aims to investigate why researchers dichotomize and assess the validity of those justifications. The authors surveyed 66 researchers and 53 Psychological Methods authors for justifications, categorized them into three groups, and examined them through logic and Monte Carlo simulations. Continuous indicators were generally superior and never substantially worse, though simulations identified specific contexts where dichotomized indicators matched or outperformed them, and other justifications were addressed with alternative techniques.

Abstract

Despite many articles reporting the problems of dichotomizing continuous measures, researchers still commonly use this practice. The authors' purpose in this article was to understand the reasons that people still dichotomize and to determine whether any of these reasons are valid. They contacted 66 researchers who had published articles using dichotomized variables and obtained their justifications for dichotomization. They also contacted 53 authors of articles published in Psychological Methods and asked them to identify any situations in which they believed dichotomized indicators could perform better. Justifications provided by these two groups fell into three broad categories, which the authors explored both logically and with Monte Carlo simulations. Continuous indicators were superior in the majority of circumstances and never performed substantially worse than the dichotomized indicators, but the simulations did reveal specific situations in which dichotomized indicators performed as well as or better than the original continuous indictors. The authors also considered several justifications for dichotomization that did not lend themselves to simulation, but in each case they found compelling arguments to address these situations using techniques other than dichotomization.