Cluster-based permutation tests are widely used in neuroscience studies for the analysis of high-dimensional electroencephalography (EEG) and event-related potential (ERP) data as it may address the multiple comparison problem without reducing the statistical power. However, classical cluster-based permutation analysis relies on parametric t-tests, whose assumptions may not be verified in case of non-normality of the data distribution and alternative options may be considered. To overcome this limitation, here we present a new software for a cluster permutation analysis for EEG series based on non-parametric Wilcoxon–Mann–Whitney tests. We tested both t-test and non-parametric Wilcoxon implementations in two independent datasets of ERPs and EEG spectral data: while t-test-based and non-parametric Wilcoxon-based cluster analyses showed similar results in case of ERP data, the t-test implementation was not able to find clustered effects in case of spectral data. We encourage the use of non-parametric statistics for a cluster permutation analysis of EEG data, and we provide a publicly available software for this computation.
Cluster permutation analysis for EEG series based on non-parametric Wilcoxon–Mann–Whitney statistical tests
Valenza G.
2022-01-01
Abstract
Cluster-based permutation tests are widely used in neuroscience studies for the analysis of high-dimensional electroencephalography (EEG) and event-related potential (ERP) data as it may address the multiple comparison problem without reducing the statistical power. However, classical cluster-based permutation analysis relies on parametric t-tests, whose assumptions may not be verified in case of non-normality of the data distribution and alternative options may be considered. To overcome this limitation, here we present a new software for a cluster permutation analysis for EEG series based on non-parametric Wilcoxon–Mann–Whitney tests. We tested both t-test and non-parametric Wilcoxon implementations in two independent datasets of ERPs and EEG spectral data: while t-test-based and non-parametric Wilcoxon-based cluster analyses showed similar results in case of ERP data, the t-test implementation was not able to find clustered effects in case of spectral data. We encourage the use of non-parametric statistics for a cluster permutation analysis of EEG data, and we provide a publicly available software for this computation.File | Dimensione | Formato | |
---|---|---|---|
PIIS2352711022001042.pdf
accesso aperto
Tipologia:
Versione finale editoriale
Licenza:
Creative commons
Dimensione
1.23 MB
Formato
Adobe PDF
|
1.23 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.