Complete effect decomposition for an arbitrary number of multiple ordered mediators with time-varying confounders: A...

Area of Science:

• Causal inference
• Statistical modeling
• Data science

Background:

• Causal mediation analysis is crucial for understanding mechanisms.
• Traditional path-specific effects are often unidentifiable with multiple mediators.
• Existing methods are limited to a small number of mediators.

Purpose of the Study:

• To generalize path-specific effects for multiple causally ordered mediators.
• To address the non-identifiability of path-specific effects using an interventional analog.
• To develop a general approach for complex causal structures with time-varying confounders.

Main Methods:

• Generalized definition of traditional and interventional path-specific effects.
• Recursive formula for nonparametric identification.
• Development of a general approach for an arbitrary number of mediators and time-varying confounders.

Main Results:

• Identifiable path-specific effects in settings with multiple mediators.
• A recursive formula for calculating these effects.
• A generalizable method applicable to complex causal structures.

Conclusions:

• The proposed methods enhance causal effect decomposition.
• This work aids in disentangling complex causal mechanisms.
• The developed approach supports data-driven scientific discovery.