Single World Intervention Graphs (SWIGs): A Practical Guide

Area of Science:

• Epidemiology
• Causal Inference
• Statistical Modeling

Background:

• The potential outcomes approach is fundamental to modern causal inference.
• Directed Acyclic Graphs (DAGs) have limitations in explicitly representing potential outcomes.
• Visualizing complex causal relationships requires advanced graphical tools.

Purpose of the Study:

• To provide a user-friendly guide for visualizing causal theories using Single World Intervention Graphs (SWIGs).
• To demonstrate how SWIGs integrate the potential outcomes framework with causal diagrams.
• To offer practical insights for constructing SWIGs in various causal scenarios.

Main Methods:

• Introduction of SWIGs as a tool to extend DAGs for causal inference.
• Step-by-step guidance on transforming DAGs into SWIGs.
• Illustrative examples of SWIG construction for confounding, mediation, and sequential randomization.

Main Results:

• SWIGs provide a clear visualization of causal assumptions and estimands.
• The g-formula can be visually identified and understood using SWIGs.
• SWIGs facilitate explicit integration of potential outcomes into causal diagrams.

Conclusions:

• SWIGs offer a powerful and intuitive method for visualizing causal inference.
• This approach enhances the application of the potential outcomes framework in research.
• Epidemiologists and researchers can expand their causal inference toolkit with SWIGs.