Hypothesis Tests of Indirect Effects for Multiple Mediators

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Mediation analysis assesses direct vs. indirect effects of independent variables.
• Single mediator models are often insufficient for complex data.
• High-dimensional data necessitates advanced mediation analysis techniques.

Purpose of the Study:

• Propose novel methods for testing indirect effects with multiple mediators and interactions.
• Address limitations of existing mediation analysis approaches.
• Incorporate correlated path effect estimators and confidence interval usage.

Main Methods:

• Development of new statistical tests for multiple mediator and interaction effects.
• Allowing for correlated estimators of path effects.
• Utilizing confidence intervals to assess the significance of mediation effects.

Main Results:

• Proposed methods demonstrate robust performance in simulation studies.
• Comparison with existing methods highlights the advantages of the new approach.
• Successful application to real-world data from the CARDIA study.

Conclusions:

• The new methods provide a more comprehensive approach to mediation analysis.
• These techniques are valuable for understanding complex indirect effects in research.
• The study enhances the toolkit for analyzing mediation with multiple mediators and interactions.