The effects of EEG data transformations on the solution accuracy of principal component analysis

Abstract

Principal component analysis (PCA) is a commonly used multivariate procedure that reduces the dimensionality of a data set. When applied to quantitative electroencephalogram (qEEG) data, PCA produces components that may represent functional systems within the brain. Unfortunately, qEEG, like many other physiological measures, produce distributions that are positively skewed. In response, researchers often transform qEEG data prior to conducting a PCA, which does not require univariate or multivariate normality. Despite this, researchers continue to transform qEEG data with limited knowledge of how such transformations will affect the accuracy (precision) of their component solutions. The purpose of the present investigation was to examine the effects of several commonly used data transformation procedures on PCA solution accuracy.