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

Challenges in interpreting results from 'multiple regression' when there is interaction between covariates.

Ian Shrier1,2, Annabelle Redelmeier3, Mireille E Schnitzer4

  • 1Epidemiology, Lady Davis Institute for Medical Research, Montreal, Québec, Canada ian.shrier@mcgill.ca.

BMJ Evidence-Based Medicine
|August 24, 2019
PubMed
Summary

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

Re-establishing Confidence in Confidence Intervals: An Evaluation of Recent Practices in Sport Injury Epidemiology.

Sports medicine (Auckland, N.Z.)·2026
Same author

The risk of subsequent concussion in adolescent ice hockey players with ≥2 concussions.

Journal of science and medicine in sport·2026
Same author

Early Detection Intervals for Evaluating Event-Based Surveillance System: Reference Dataset Development Study.

JMIR public health and surveillance·2026
Same author

Identification of the causal odds ratio in test negative designs.

International journal of epidemiology·2026
Same author

Introducing a new "Preliminary Report" submission category for small-sample intervention studies: viewpoints from external experts.

Science & medicine in football·2026
Same author

Adaptive Sparsening and Smoothing of the Treatment Model for Longitudinal Causal Inference Using Outcome-Adaptive LASSO and Marginal Fused LASSO.

Statistics in medicine·2026

Interpreting regression results requires caution. Adjusted effects in studies are often misinterpreted as population average causal effects (PACEs) when interactions are present, necessitating correct calculation methods.

Area of Science:

  • Biostatistics
  • Epidemiology
  • Medical Research Methodology

Background:

  • Accurate interpretation of research findings is crucial for evidence-based medicine.
  • Adjusted effect estimates from regression analyses are commonly reported in observational studies and randomized trials.
  • These estimates are often misinterpreted as population average causal effects (PACEs).

Purpose of the Study:

  • To demonstrate that the standard interpretation of adjusted regression effects as PACEs is incorrect when treatment interactions exist.
  • To present correct methods for calculating PACEs from regression analyses.
  • To discuss alternative causal inference methods and the importance of considering interaction terms.

Main Methods:

  • Review of statistical interpretation of regression models in causal inference.
Keywords:
conditional estimatesinverse probability treatment weightingmarginal estimatesregression

Related Experiment Videos

  • Derivation of correct methods for estimating PACEs in the presence of interactions.
  • Discussion of alternative causal inference techniques.

    • Main Results:

    • The common interpretation of adjusted regression effects as PACEs is flawed when interactions between the exposure and other variables are present.
    • Correct methods for calculating PACEs from regression models are provided.
    • The significance of interaction terms in regression analyses is highlighted.

    Conclusions:

    • Researchers must use appropriate methods to calculate population average causal effects (PACEs) when interaction terms are present in regression models.
    • Excluding interaction terms based solely on p-values can lead to incorrect causal effect estimates.
    • Correct causal inference methodology is essential for advancing evidence-based medicine.