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

Identifiability and exchangeability for direct and indirect effects.

J M Robins1, S Greenland

  • 1Occupational Health Program, Harvard School of Public Health, Boston, MA 02115.

Epidemiology (Cambridge, Mass.)
|March 1, 1992
PubMed
Summary

Estimating direct and indirect effects of exposures is complex. Common adjustment methods can be biased, and separating these effects requires specific assumptions or trial designs, often necessitating the G-computation algorithm.

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

On the asymptotic validity of confidence sets for linear functionals of solutions to integral equations.

Biometrika·2025
Same author

Using negative controls to identify causal effects with invalid instrumental variables.

Biometrika·2025
Same author

Multiple robustness in factorized likelihood models.

Biometrika·2018
Same author

Instrumental variables as bias amplifiers with general outcome and confounding.

Biometrika·2017
Same author

Marginal Mean Models for Dynamic Regimes.

Journal of the American Statistical Association·2009
Same author

Glaucoma outcome studies using existing databases: opportunities and limitations.

Journal of glaucoma·2009

Area of Science:

  • Epidemiology
  • Biostatistics
  • Causal Inference

Background:

  • Separating direct and indirect effects of exposures is crucial for understanding disease.
  • Traditional adjustment methods for intermediate variables may introduce bias.
  • Identifiability of direct and indirect effects is a key challenge in causal inference.

Purpose of the Study:

  • To investigate the identifiability and estimation of direct and indirect effects.
  • To evaluate the bias associated with conventional adjustment methods.
  • To explore conditions and methods for separating causal effects.

Main Methods:

  • Theoretical analysis of causal effect decomposition.
  • Simulation studies under various interaction and randomization scenarios.

Related Experiment Videos

  • Application of the G-computation algorithm for estimation.
  • Consideration of randomized trials and observational data.

    • Main Results:

    • Adjustment for intermediate variables is often biased for direct effect estimation.
    • Direct and indirect effects are not separately identifiable in standard randomized exposure trials.
    • Separation is possible with dual randomization (exposure and intervention) and G-computation, provided no interaction.
    • The fraction of disease preventable by controlling intermediates can be estimated even without dual randomization.

    Conclusions:

    • Conventional methods for estimating direct effects are unreliable.
    • Special assumptions or advanced methods like G-computation are required for valid effect separation.
    • Careful consideration of interactions and trial design is essential for causal effect decomposition.