Testing Correlated "BEG-Like" Data for Normality Using a Modified Kolmogorov-Sminov Statistic
Marc S. Weiss · IEEE Transactions on Biomedical Engineering · 1986
The one-sample Kolmogorov-Smirnov goodness-of-fit test (KS) is designed for use with independent data and can be highly sensitive to correlated data. Standard critical values for KS cannot be used with data with known correlations. For data with EEG-like spectra (low frequency, high amplitude spectral peaks) an empirically derived correction for KS provides correct critical values and retains much of the power of the original KS. The. correction is based on a simple quadratic expression involving a parameter τ computed from zero-crossing measurements.