Intergroup and intrasubject principal component analysis of event-related potentials

The purpose of this paper is to show that the familiar principal component analysis (PCA) of event-related potentials is identical to an easily formulated least squares method. This correspondence permits interpretation of several criticisms of PCA and clearer presentation of its strengths and shortcomings. Because data analysis based on PCA compares amplitudes of empirically derived components, it is necessary that the shape of the components be similar under the experimental conditions. We present and illustrate a statistical method for comparison of principal components across groups and between conditions within subjects.