Projection onto centroids difference vectors: a new approach to determine between group topographical differences, applied to P3 amplitude in schizophrenia.

A common problem in electrophysiological research concerns between group analysis of EEG and ERP topography. This paper proposes a new technique for determining whether or not a significant difference exists between multiple scalp site measurements from two groups. The method requires no a priori assumptions about the data and is thus ideal for exploratory data analysis, and it also requires that only one statistical test need be performed (significantly reducing the possibility of type I errors). The Projection onto Centroids Difference Vectors (PCDV) method involves deriving a measure from each individual of whether their measurements across sites are closer overall to the mean measurements of the rest of their experimental group, or to the other group. These measures from each individual are then compared between groups using a Student's t test, which indicates whether one group's data is significantly spatially different from the other. In this study we describe the method in detail and apply it to both stimulated data and to real auditory P3 data in unmedicated, medicated schizophrenics and matched normal controls. The PCDV method was also compared with statistical probability mapping (SPM). The PCDV method revealed the differences between the normal and patient groups more unambiguously than SPM, and the simulated data revealed that it was not liable to type I errors. PCDV provides an appropriate method for testing any between group EEG and ERP topographical differences.