Repeated measures analysis for steady-state evoked potentials

Area of Science:

• Neuroscience
• Statistics
• Signal Processing

Background:

• Brain responses to repetitive stimuli yield steady-state evoked potentials (ssEP).
• Analyzing ssEP Fourier measurements requires statistical methods to differentiate neural responses across experimental conditions within participants.
• Existing methods may struggle with complex data structures and assumptions.

Purpose of the Study:

• Introduce novel statistical methods for comparing multiple dependent clusters of Fourier measurements from ssEP data.
• Address the analysis of neural responses across various experimental conditions within the same individuals.

Main Methods:

• Developed two statistics: one based on repeated measures ANOVA for complex numbers (circular clusters, equal variance).
• A second statistic, derived from the rank-sum Friedman test, handles violations of circularity or equal variance assumptions (elliptical clusters).

Main Results:

• Validated statistics using simulated and empirical ssEP data.
• Methods maintain a constant Type-I error rate of 0.05 across conditions, including unequal variance-covariance matrices and outliers.
• Demonstrated lower Type-II error rates compared to repeated measures multivariate analysis of variance (rmMANOVA).

Conclusions:

• The new statistical methods effectively compare multiple dependent Fourier estimate clusters across experimental conditions.
• These tools are robust to unequal variances and outliers in ssEP data.
• Enable more reliable differentiation of neural responses in complex experimental designs.