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

Assessing interaction in case-control studies: type I errors when using both additive and multiplicative scales.

Jacqueline R Starr1, Barbara McKnight

  • 1Department of Pediatrics, University of Washington, Children's Hospital and Regional Medical Center, Childrend's Craniofacial Center, Seattle 98105-0371, USA. jrstarr@u.washington.edu

Epidemiology (Cambridge, Mass.)
|July 3, 2004
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

Inference for Stationary Log-Gaussian Cox Point Processes using Bayesian Deep Learning: Application to Human Oral Microbiome Image Data.

Spatial statistics·2026
Same author

Building process improvement capacity in epidemiologic research operations: The Nurses' Health Studies experience.

American journal of epidemiology·2026
Same author

Effect of Replacing Added Sugars with Sucralose on Gut Microbiome Composition Among Asian Indian Adults in Two 12-week Randomized Controlled Trials.

Current developments in nutrition·2025
Same author

Building process improvement capacity in epidemiologic research operations: The Nurses' Health Study experience.

medRxiv : the preprint server for health sciences·2025
Same author

Group-wise normalization in differential abundance analysis of microbiome samples.

BMC bioinformatics·2025
Same author

Inference for Log-Gaussian Cox Point Processes using Bayesian Deep Learning: Application to Human Oral Microbiome Image Data.

ArXiv·2025
Same journal

Application of the E-value under non-proportional hazards.

Epidemiology (Cambridge, Mass.)·2026
Same journal

Can the All of Us sample be reweighted to mirror a nationally representative sample? A comparison of mortality predictors.

Epidemiology (Cambridge, Mass.)·2026
Same journal

Gut health, systemic inflammation, and linear growth among Indonesian infants: findings from the Action Against Stunting Hub observation cohort: Erratum.

Epidemiology (Cambridge, Mass.)·2026
Same journal

Evaluating Estimators in Partially Identified Models.

Epidemiology (Cambridge, Mass.)·2026
Same journal

Stratification and accumulation? Explaining changing mortality inequities between business owners and non-owners in the U.S. (1984-2022).

Epidemiology (Cambridge, Mass.)·2026
Same journal

Be wary of age-stratum aging in early-onset cancer trends.

Epidemiology (Cambridge, Mass.)·2026
See all related articles

When assessing effect modification in epidemiologic studies, using both multiplicative (M) and additive (A) scales can inflate false-positive findings. Researchers should interpret interaction tests cautiously, especially when the analysis scale is not pre-specified.

Area of Science:

  • Epidemiology
  • Biostatistics

Background:

  • Epidemiologic studies frequently assess effect modification using multiple statistical scales.
  • This dual approach, employing both multiplicative (M) and additive (A) scales, may increase the likelihood of false-positive interaction findings.

Purpose of the Study:

  • To investigate the extent of Type I error inflation when evaluating statistical interactions using both multiplicative and additive scales.
  • To provide empirical evidence on the validity of common effect modification assessment approaches in case-control studies.

Main Methods:

  • Computer simulations were conducted to assess Type I error rates.
  • Evaluated the combined use of multiplicative interaction coefficients (M) from logistic regression and additive interaction coefficients (A) from additive relative risk regression.

Related Experiment Videos

Main Results:

  • The overall Type I error rate frequently exceeded the 5% threshold when both multiplicative and additive interaction tests were performed concurrently.
  • Empirical evidence suggests limitations in the validity of simultaneously assessing effect modification on different statistical scales.

Conclusions:

  • Simultaneous testing for effect modification on both multiplicative and additive scales can lead to inflated Type I error rates.
  • Interaction hypothesis tests for effect modification require cautious interpretation, particularly when the analysis scale is not determined a priori.
  • Selecting an interaction test based solely on the lowest P-value is not a scientifically justified approach.