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

PRESS-related statistics: regression tools for cross-validation and case diagnostics

D B Holiday1, J E Ballard, B C McKeown

  • 1Department of Epidemiology/Biomathematics, University of Texas Health Center at Tyler 75710, USA.

Medicine and Science in Sports and Exercise
|April 1, 1995
PubMed
Summary

Data-splitting for regression validation has drawbacks. The PRESS statistic offers an alternative, providing unbiased performance estimates without splitting data, especially useful for smaller datasets in health science research.

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

Hepatic metabolism of MK-0457, a potent aurora kinase inhibitor: interspecies comparison and role of human cytochrome P450 and flavin-containing monooxygenase.

Drug metabolism and disposition: the biological fate of chemicals·2007
Same author

Asbestos in extrapulmonary sites: omentum and mesentery.

Chest·2000
Same author

The Bruce treadmill protocol: does walking or running during the fourth stage alter oxygen consumption values?

The Journal of sports medicine and physical fitness·1998
Same author

Assessing influenza immunization rates in Medicare managed care plans: a comparison of three methods.

The Joint Commission journal on quality improvement·1997
Same author

Expanding use of thrombolytic therapy in the treatment of acute myocardial infarction in rural Alaskan hospitals.

Alaska medicine·1997
Same author

Beta-Carotene, vitamin A, and lung cancer chemoprevention: results of an intermediate endpoint study.

The American journal of clinical nutrition·1995

Area of Science:

  • Health Sciences
  • Biostatistics
  • Regression Analysis

Background:

  • Data-splitting is a common regression validation method in health sciences.
  • This technique uses a fitting sample for model development and a validation sample for performance estimation.
  • While it corrects for upward bias, data-splitting has practical limitations.

Purpose of the Study:

  • To contrast data-splitting with the PRESS statistic for regression equation validation.
  • To highlight the advantages of the PRESS statistic, particularly for smaller datasets.
  • To recommend reporting methods for these validation statistics.

Main Methods:

  • The study simulated data for predicting body density from anthropometric measurements.
  • It compared the traditional data-splitting method with the PRESS statistic.

Related Experiment Videos

  • A dataset of 117 women was used for the simulation and comparison.

    • Main Results:

    • Data-splitting presents challenges including sample matching difficulties and reduced equation stability.
    • The PRESS statistic provides unbiased estimates of R2 and SEE without data splitting.
    • The PRESS method also offers valuable case diagnostics and is readily available in statistical software.

    Conclusions:

    • The PRESS statistic is a superior alternative to data-splitting for validating regression equations.
    • It is particularly advantageous for smaller datasets common in health science research.
    • Wider adoption of the PRESS statistic is recommended for robust regression model validation.