Testing EEG data for statistical normality

A modification of the Kolmogorov-Smirnov (KS) statistical test for normally distributed time series data when these data are highly correlated and EEG-like is described. Analysis of EEG data using the KS statistics or any other statistic that needs independent data requires reduction of the effective sampling rate by discarding data. Here an estimate of the correlation properties of the data is used to correct the test statistic, allowing all the data to be used as originally sampled. This increases the power of the test. Results obtained when the KS procedure was applied to several samples of EEG data obtained under various experimental conditions are reported. They demonstrate the usefulness of the modified method as a test for normality of correlated EEG data with strong peaks in the lower part of the spectrum.<>