The cost of dichotomising continuous variables

By Douglas G. Altman, Patrick Royston. British medical journal, 2006

2,417 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/61e0e909828fdab8161292b40e0cecf6edaeceb5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Dominance and Age in Bilingualism. (2014, background): “With continuous variables generally, regression 35 models also eliminate the arbitrariness associated with stipulating group mem- bership and offer more statistical power in analyses (Nelson and Zaichkowsky 1979; Altman and Royston 2006).”
A researcher's guide to regression, discretization, and median splits of continuous variables (2015, background): “…effect size estimates Cox (1957), Gelman and Park (20 Misrepresentation of continuous constructs as fictional groups MacCallum et al. (2002) Concealment of nonlinearity Altman and Royston (2006), Mac reduced predictive performance, increased Type II error, and inefficient effect size estimates.”
Consumer motivations for sustainable consumption : the interaction of gain, normative and hedonic motivations on electric vehicle adoption (2018, background): “Dichotomizing continuous variables, as we did with perceived social norms, can results in some information loss (Altman & Royston, 2006), however it is done for practicality and simplicity of interpretation.”
Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies: The CHARMS Checklist (2014, methodology): “, using restricted cubic splines or fractional polynomials) are indicated [45,51,58,93,94,96].”
Introducing the Open Affective Standardized Image Set (OASIS) (2017, result): “The dichotomization of continuous variables is useful for ease of understanding, but it produces a loss of power and can lead to spurious results (Altman & Royston, 2006; MacCallum, Zhang, Preacher, & Rucker, 2002).”
Developing clinical prediction models: a step-by-step guide (2024, unknown): “38 Step 4: Identify candidate predictors and specify measurement methods”
Emotion-regulation strategies across psychopathology: A meta-analytic review. (2010, background): “…dichotomize the sample, we favored the continuous (i.e. correlational) over the dichotomized data, because splitting a variable into categories results in loss of information and might increase the probability of type II errors (Altman & Royston, 2006; Rosenthal & DiMatteo, 2001; Streiner, 2002).”
Informative tools for characterizing individual differences in learning: Latent class, latent profile, and latent transition analysis (2017, background): “…abilities on learning outcomes, the use of arbitrary cutoff points is considered poor statistical practice: it is a-theoretical and introduces error which can result in a distorted picture of relations between variables (Altman & Royston, 2006; Irwin & McClelland, 2003; Maxwell & Delaney, 1993).”
GRADE guidelines: 13. Preparing summary of findings tables and evidence profiles-continuous outcomes. (2013, background): “In deciding whether to dichotomize the outcome, review authors should also consider possible loss of statistical power [21].”
Prognosis and prognostic research: Developing a prognostic model (2009, background): “Often, new variables need to be created (for example, diastolic and systolic blood pressure may be combined to give mean arterial pressure). For ordered categorical variables, such as stage of disease, collapsing of categories or a judicious choice of coding may be required. We advise against turnin”
2016 Revisions to the 2010/2011 fibromyalgia diagnostic criteria. (2016, background): “In addition, the nature of the fibromyalgia symptom continuum makes differences in severity at the borderline of diagnosis problematic, for patients on different sides of the cut point are much more similar than they are different [35].”