Graphical representation of multidimensional EEG data and classificatory aspects.

EEG data are intrinsically multidimensional and this applies also to a set of broad band spectral parameters (6 parameters per derivation). Non-metric multidimensional scaling (MDS) and principal component analysis (PCA) are compared with respect to their capacity to achieve an adequate few-dimensional representation for the graphical comparison of groups and the identification of subgroups. In an application to neurophysiological aspects of mental retardation and learning disability, MDS proved to be somewhat superior to PCA. Both methods profited from the introduction of (1-p)-convex hulls to define a normative region.