Brain networks construction using Bayes FDR and average power function.

Area of Science:

• Neuroscience
• Graph Theory
• Statistical Analysis

Background:

• Brain functional connectivity is crucial in neuroscience, often studied using graph theory.
• Graph construction relies on adjacency matrices derived from time series, with thresholding correlation coefficients being a key step.
• Current methods for threshold selection lack robustness and can be sensitive to statistical errors.

Purpose of the Study:

• To propose a robust multiple testing approach for selecting the optimal threshold (τ) in brain connectivity analysis.
• To balance Type I and Type II statistical errors using Bayes false discovery rate and a novel average power function.
• To enhance the stability and reliability of functional connectivity estimations.

Main Methods:

• Developed a multiple testing framework incorporating Bayes false discovery rate.
• Introduced a new estimator for statistical power: the average power function.
• Utilized bootstrap distributions of the average power function and False Discovery Rate for robust threshold selection (τ).

Main Results:

• The proposed average power function estimator demonstrates reliable performance under various simulated dependence conditions.
• The method effectively balances statistical errors, leading to more stable threshold choices.
• The approach was successfully applied to functional magnetic resonance imaging (fMRI) and high-density electroencephalogram (EEG) data.

Conclusions:

• The novel multiple testing approach provides a statistically sound and robust method for threshold selection in brain connectivity studies.
• This method improves the reliability of functional connectivity measures derived from neuroimaging data.
• The proposed techniques offer a significant advancement for analyzing complex brain networks.