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

Estimating equations for measures of association between repeated binary responses

S R Lipsitz1, G M Fitzmaurice

  • 1Department of Biostatistics, Harvard School of Public Health, Boston Massachusetts 02115, USA.

Biometrics
|September 1, 1996
PubMed
Summary

This study introduces a new method using conditional residuals to estimate correlations in repeated binary responses. This approach offers improved efficiency, especially with missing data or unequal cluster sizes, compared to standard methods.

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

Reducing surgical mortality in Scotland by use of the WHO Surgical Safety Checklist.

The British journal of surgery·2019
Same author

Minimally invasive surgery and its impact on 30-day postoperative complications, unplanned readmissions and mortality.

The British journal of surgery·2017
Same author

Physician Perceptions and Beliefs about Generating and Providing a Clinical Summary of the Office Visit.

Applied clinical informatics·2015
Same author

Prenatal protein level impacts homing behavior in Long-Evans rat pups.

Nutritional neuroscience·2015
Same author

Physician beliefs about the impact of meaningful use of the EHR: a cross-sectional study.

Applied clinical informatics·2014
Same author

Prediction of time-to-attainment of recovery for borderline patients followed prospectively for 16 years.

Acta psychiatrica Scandinavica·2014

Area of Science:

  • Biostatistics
  • Statistical Methods
  • Longitudinal Data Analysis

Background:

  • Moment-based methods analyze repeated binary responses using marginal odds ratios.
  • Generalized estimating equations (GEE) with conditional residuals estimate marginal odds ratios.
  • Existing methods may lack efficiency for certain association measures or data structures.

Purpose of the Study:

  • To extend the use of conditional residuals for estimating other association measures, specifically correlation, in repeated binary data.
  • To compare the efficiency of correlation estimators based on conditional residuals against existing methods.
  • To evaluate the performance of these estimators under conditions of missing data and unequal cluster sizes.

Main Methods:

  • Utilized conditional residuals, defined as deviations about conditional expectations.

Related Experiment Videos

  • Developed estimators for correlation based on these conditional residuals.
  • Compared the efficiency of the proposed correlation estimator with maximum likelihood, GEE2, and standard GEE estimators.
  • Assessed performance in scenarios with incomplete responses and unequal cluster sizes.

    • Main Results:

    • Conditional residuals can effectively estimate correlations between paired binary responses.
    • The conditional residual-based correlation estimator shows near-maximal efficiency, except for large correlation values.
    • This estimator is more efficient than the standard GEE estimator using unconditional residuals.
    • Significant efficiency gains are observed with missing data or unequal cluster sizes.

    Conclusions:

    • Conditional residuals provide a flexible tool for estimating various association measures beyond the marginal odds ratio.
    • The proposed method offers a robust and efficient alternative for analyzing repeated binary data, particularly in challenging data situations.
    • This enhances the utility of generalized estimating equations in biostatistical research.