New methods for multiple testing in permutation inference for the general linear model.

Area of Science:

• Neuroimaging
• Statistical modeling
• Functional data analysis

Background:

• Permutation methods are standard for testing general linear models (GLMs) in neuroimaging.
• Current methods often use the maximum test statistic, which can be suboptimal.

Purpose of the Study:

• To introduce novel multiple testing procedures for GLMs in functional imaging.
• To improve statistical power and robustness compared to existing methods.

Main Methods:

• Developed new multiple testing methods based on sorting permuted functional test statistics.
• Utilized pointwise rank measures for improved performance.
• Ensured methods are applicable to large datasets.

Main Results:

• Proposed methods demonstrate improved power and robustness against test statistic inhomogeneity.
• Performance validated through simulation experiments and real brain imaging data.
• The R package GET facilitates the implementation of these procedures.

Conclusions:

• The new permutation-based multiple testing methods offer a more powerful and robust alternative for GLM analysis in neuroimaging.
• The GET package provides a practical tool for applying these advanced statistical techniques.