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

Causal conclusions are most sensitive to unobserved binary covariates.

Liansheng Wang1, Abba M Krieger

  • 1GlaxoSmithKline, 1600 Vine Street, 3F0415, Philadelphia, PA 19102, USA. liansheng.wang@wharton.upenn.edu

Statistics in Medicine
|October 13, 2005
PubMed
Summary

Unmeasured confounding in matched pairs is most sensitive to binary covariates, not long-tailed ones. Assuming a binary unobserved covariate provides a conservative estimate of treatment-response bias.

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

Multi-omics Signature Predicts Anti-EGFR Therapy Benefit in Colorectal Cancer Liver Metastases: A Multi-center Cohort Study.

Current cancer drug targets·2026
Same author

Finerenone combined with guideline-directed medical therapy in patients with post-myocardial infarction heart failure: a real-world study.

Frontiers in pharmacology·2026
Same author

Deep learning predicts stent implantation in borderline coronary lesions from angiography.

NPJ digital medicine·2026
Same author

M2OTCA: Multiple-magnification optimal transport-based cross-attention learning for whole slide image classification.

Medical image analysis·2026
Same author

MorphoNet: Morphological sub-region-based structure learning for WSI analysis.

Medical image analysis·2026
Same author

Real-world performance of open-source large language models in diabetes diagnosis.

Frontiers in endocrinology·2026

Area of Science:

  • Statistics
  • Biostatistics
  • Epidemiology

Background:

  • Unobserved covariates can confound the relationship between treatment and response.
  • Quantifying unmeasured bias typically assumes the covariate is binary or within the unit interval.

Purpose of the Study:

  • To relax the assumption on unobserved covariate distribution in matched pairs with binary outcomes.
  • To investigate the impact of different unobserved covariate distributions on bias quantification.

Main Methods:

  • Analysis of matched pairs with binary treatment and response.
  • Relaxing the typical assumption of a binary or unit-interval unobserved covariate.

Main Results:

  • Contrary to intuition, binary unobserved covariates cause the most bias, making this assumption the most conservative.

Related Experiment Videos

  • A normally distributed unobserved covariate leads to less bias sensitivity compared to a binary one.
  • Treatment-response relationships sensitive to dichotomous unobserved covariates may become insensitive with normally distributed covariates.

    • Conclusions:

    • For matched pairs with binary treatment/response, assuming a binary unobserved covariate is a conservative approach to assessing bias.
    • Alternative covariate distributions, like normal, may show less sensitivity to unmeasured bias than commonly assumed binary ones.
    • Findings are specific to the matched pair setting and warrant further investigation in other contexts.