Sparse Principal Component based High-Dimensional Mediation Analysis.

Area of Science:

• Statistics
• Biostatistics
• Neuroimaging Analysis

Background:

• Causal mediation analysis quantifies intermediate effects in treatment-outcome pathways.
• Multiple, dependent mediators pose challenges due to exponential pathway decomposition.
• Existing principal component analysis (PCA) methods lack interpretability.

Purpose of the Study:

• To propose a sparse high-dimensional mediation analysis approach using sparse PCA.
• To improve the interpretability of mediator effects in complex causal pathways.
• To apply the novel method to functional magnetic resonance imaging (fMRI) data.

Main Methods:

• Developed a sparse PCA method tailored for causal mediation analysis.
• Applied the sparse PCA approach to a task-based fMRI dataset.
• Evaluated the method's ability to identify biologically relevant mediators.

Main Results:

• The proposed sparse PCA method offers improved interpretability over standard PCA.
• The approach successfully identified biologically meaningful mediators in the fMRI study.
• Demonstrated the utility of sparse high-dimensional mediation analysis.

Conclusions:

• Sparse PCA provides a powerful and interpretable tool for high-dimensional causal mediation analysis.
• This method advances the analysis of complex biological pathways, particularly in neuroimaging.
• The approach facilitates the discovery of interpretable mediators in complex datasets.