A novel causal mediation analysis approach for zero-inflated mediators

Area of Science:

• Biomedical Research
• Causal Inference
• Statistical Modeling

Background:

• Mediation analysis is crucial for causal inference in biomedical research.
• Existing mediation frameworks struggle with zero-inflated mediators due to excessive zeros.
• True and false zeros in mediators pose significant analytical challenges.

Purpose of the Study:

• To develop a novel mediation modeling approach for zero-inflated mediators.
• To address challenges posed by both true and false zeros in mediation analysis.
• To decompose the total mediation effect into components specific to zero-inflated structures.

Main Methods:

• Developed a novel mediation modeling approach tailored for zero-inflated mediators.
• Decomposed the total mediation effect into two distinct components.
• Conducted extensive simulation studies to evaluate the approach's performance.
• Applied the proposed method to a real-world biomedical study.

Main Results:

• The proposed approach effectively handles mediators with zero-inflated structures.
• The novel method decomposes mediation effects into numerical and binary change components.
• Simulation results demonstrate superior performance compared to standard causal mediation analysis.
• The approach showed practical utility in a real study application.

Conclusions:

• The novel mediation modeling approach successfully addresses zero-inflated mediators.
• This method offers a significant advancement over existing techniques for complex mediation scenarios.
• The approach provides a more nuanced understanding of causal pathways involving zero-inflated variables.