Finally, because EEG measures typically are collected across different conditions at frequent intervals, the measures tend to be highly correlated, especially for adjacent conditions. This violates the sphericity assumption of repeated-measures ANOVA and can lead to invalid statistical results. Thus, researchers should use Huynh-Feldt, Geisser-Greenhouse, or other similar corrections to alter the degrees of freedom and the significance of the F-ratio. The epsilon value should be reported in manuscripts (Jennings, 1987). An appropriate alternative is to analyze data with repeated-measures MANOVA, which does not assume sphericity. No corrections are required when using MANOVA to analyze EEG data.
