The Practical Effect of Routine Data Transformations on Absolute EEG Power Derived from Spectral Analysis

This investigation was conducted to evaluate the effects of selected EEG data transformations on analysis of variance ANOVA results in a repeated measures design. Eighteen subjects were given resting EEG evaluations at 8 different times during a period of continuous wakefulness. Following spectral analysis, the data were either analyzed as untransformed absolute EEG power, log-natural transformed power, or 2-arcsine-square root transformed relative power. Results indicated that while the relative power transformation lead to a more sensitive statistical analysis, it concurrently introduced data interpretation problems. In contrast, the results with both untransformed and transformed absolute power were quite similar. Overall, it was concluded that although transformations improve the Gaussian properties of the data, they do not appear to substantially impact the conclusions that will be drawn from a repeated measures.