Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Consequences of dichotomization.

Valerii Fedorov1, Frank Mannino, Rongmei Zhang

  • 1Research Statistics Unit, Biomedical Data Sciences, GlaxoSmithKline Pharmaceuticals, Collegeville, PA 19426, USA.

Pharmaceutical Statistics
|April 5, 2008
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Stroke and Bleeding Risks With Non-Vitamin K Oral Anticoagulants in Nonvalvular Atrial Fibrillation.

JAMA network open·2026
Same author

Self-efficacy mediates social support and self-management behaviors in kidney transplant recipients: A systematic review and meta-analytic structural equation modeling analysis.

Journal of psychosomatic research·2025
Same author

A study on the spatial distribution characteristics and influencing factors of forest villages in southwest China based on OPGD.

Scientific reports·2025
Same author

Self-management experiences and coping strategies of glaucoma patients: A meta-synthesis of qualitative research.

Patient education and counseling·2025
Same author

Correction: An Evaluation of Duplicate Adverse Event Reports Characteristics in the Food and Drug Administration Adverse Event Reporting System.

Drug safety·2025
Same author

An Evaluation of Duplicate Adverse Event Reports Characteristics in the Food and Drug Administration Adverse Event Reporting System.

Drug safety·2025
Same journal

Impact of Information Leakage in Platform Trials With Survival Endpoints on Type I Error Control.

Pharmaceutical statistics·2026
Same journal

Harmonic Fowlkes-Mallows Index for Medical Diagnostics Tests and Optimal Cut-Off Point Selection of Binary Diseases.

Pharmaceutical statistics·2026
Same journal

Early Phase Dose-Finding Designs for CAR-T Cell Therapies.

Pharmaceutical statistics·2026
Same journal

Optimizing Randomization Ratios in Clinical Trials With Survival Endpoints.

Pharmaceutical statistics·2026
Same journal

CUI-MET: A Clinical Utility Index Based Analysis and Decision Framework for Dose Optimization in Multiple-Dose, Multiple-Outcome Randomized Trials.

Pharmaceutical statistics·2026
Same journal

Will the Pharmaceutical Industry Need Statisticians in an AI World?

Pharmaceutical statistics·2026
See all related articles

Transforming continuous data to binary outcomes (dichotomization) significantly reduces statistical information, at least by 36% for normal data. This loss impacts hypothesis testing and requires larger sample sizes, making dichotomization generally inadvisable.

Area of Science:

  • Statistics
  • Biostatistics
  • Data Analysis

Background:

  • Dichotomization, the conversion of continuous outcomes to binary ones, is a common but statistically problematic practice.
  • This transformation can lead to substantial information loss, affecting statistical estimation and hypothesis testing accuracy.

Purpose of the Study:

  • To quantify the information loss associated with dichotomization of continuous outcomes.
  • To evaluate the statistical implications of dichotomization for estimation and power.

Main Methods:

  • Analysis of information loss using Fisher's information for normally distributed data.
  • Assessment of the impact of cut-point selection on information loss.
  • Comparison of statistical power between continuous and dichotomized data.

Related Experiment Videos

Main Results:

  • Dichotomization results in a minimum information loss of 36% for normally distributed data.
  • Information loss is highly dependent on the chosen cut points, which are often based on unknown parameters.
  • Maintaining statistical power with dichotomized data requires a larger sample size compared to using continuous data.

Conclusions:

  • Dichotomization of continuous outcomes is statistically detrimental, leading to significant information loss and reduced power.
  • The practice should be avoided in most statistical analyses, especially for estimation and hypothesis testing.
  • Exceptions may exist in specific scenarios, such as estimating cumulative distribution functions when the assumed model deviates significantly from the true model.