Flexible Multi-Step Hypothesis Testing of Human ECoG Data using Cluster-based Permutation Tests with GLMEs.

Area of Science:

• Neuroscience
• Statistics
• Computational Biology

Background:

• Cluster-based permutation tests (CBPT) are widely used for analyzing electrophysiological signals like EEG and MEG.
• Traditional CBPT has limitations in handling multiple fixed effects and random effects (e.g., inter-subject variability).
• A flexible hypothesis testing strategy is needed for diverse experimental designs and data types in brain signal analysis.

Approach:

• Developed a multi-step hypothesis testing strategy using CBPT with Linear Mixed Effects Models (LMEs) and Generalized Linear Mixed Effects Models (GLMEs).
• Evaluated the robustness of LMEs and GLMEs using simulated data.
• Applied the strategy to human ECoG data for analyzing ERPs and broadband power, and to high-gamma burst data for nonlinear analysis.

Key Points:

• LMEs and GLMEs demonstrated robustness and outperformed traditional methods in simulations, especially with suboptimal data and small effect sizes.
• CBPT with LMEs showed superior statistical power compared to separate t-tests for analyzing broadband power signals.
• The proposed method successfully analyzed human ECoG data, replicating known latency effects and handling nonlinear data.

Conclusions:

• The proposed CBPT with LMEs/GLMEs approach is robust for analyzing linear and nonlinear electrophysiological data with multiple fixed and random effects.
• This methodology enhances statistical power and accuracy, improving reproducibility in brain signal analysis.
• The approach is versatile, applicable to individual channels or group-level data for comparing functional or anatomical groups.