Reducing Bias with Directed Acyclic Graphs
Author Information
Author(s): Ian Shrier, Robert W. Platt
Primary Institution: McGill University
Hypothesis
Can a simple 6-step approach using directed acyclic graphs (DAGs) effectively reduce bias in causal inference?
Conclusion
Using a simple 6-step DAG approach is likely to reduce bias in effect estimates in statistical models.
Supporting Evidence
- The traditional approach to confounding may create bias rather than eliminate it.
- Many researchers continue to use outdated methods due to a lack of understanding of DAGs.
- The 6-step process helps identify appropriate covariates to minimize bias.
Takeaway
This study shows a way to use diagrams to help researchers avoid mistakes when figuring out if one thing causes another.
Methodology
The manuscript outlines a 6-step process for using directed acyclic graphs to assess and minimize bias in causal inference.
Potential Biases
The traditional methods of adjusting for confounding may introduce new biases instead of minimizing them.
Limitations
The study is limited to a conceptual discussion and does not provide empirical validation of the 6-step process.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website