matrix by a general loading matrix and by allowing the number of spatial and temporal components to be different (Wang et al. 2000). Thus, the trilinear model has the advantages of both SVD and TCM methods. The trilinear components are uniquely determined and more interpretable. Trilinear modeling can be used for interindividual comparison studies, single-trial modeling, clinical classification of patients, and data filtering. For example, the trilinear method was applied in “dynamic time warping” to align the repeated single trials of the ERPs in order to eliminate the timing differences and to get an improved estimate of the ERP components (Wang et al. 2001). In their original work, Wang and colleagues (2000) had demonstrated the decomposition of visual P3 into 16 spatio-temporal components. Significant linkage between time-warped P3-related trilinear components in a visual oddball paradigm in densely affected alcoholic families from the Collaborative Study on the Genetics of Alcoholism (COGA) has been reported (Porjesz et al. 2002b).
