Practical guidance for conducting mediation analysis with multiple mediators using inverse odds ratio weighting

Area of Science:

• Biostatistics
• Epidemiology
• Econometrics

Background:

• Mediation analysis is crucial for understanding causal pathways.
• Existing techniques are often complex or limited in scope.
• A need exists for accessible and versatile mediation analysis methods.

Purpose of the Study:

• To present a novel mediation analysis technique using inverse odds ratio weighting (IORW).
• To provide practical guidance for implementing IORW for natural direct and indirect effects.
• To demonstrate the applicability of IORW across various regression models.

Main Methods:

• Utilizes inverse odds ratio weighting (IORW) based on the odds ratio's invariance property.
• Weights regression analyses by regressing exposure on mediators and covariates.
• Estimates natural direct effects via weighted regression; indirect effects are derived from total effects.

Main Results:

• IORW effectively estimates natural direct and indirect effects in mediation analyses.
• The method accommodates multiple discrete or continuous mediators.
• IORW is compatible with standard regression models, including quantile and survival analysis.

Conclusions:

• IORW offers a practical, implementable, and versatile approach to mediation analysis.
• This technique simplifies the estimation of complex causal pathways.
• The method was validated using data from the Moving to Opportunity experiment.