Multiway component analysis

Multiway component analysis refers to the analysis of three- and higher way data sets, for instance, scores of a number of individuals, on a number of variables, in a number of different situations. Multiway component analysis got an important new world-wide impetus in the early nineties when psychologists started collaboration on this topic with chemists. My involvement in this led to several papers in the psychometrical and chemometrical literature, as well as contributions to organization of “TRICAP” conferences on this topic. This new impetus also inspired a reevaluation of the usability of three-way methods in the realm of psychology, as expressed in papers on enhancing the efficiency of the computational procedures, procedures to enhance interpretability of three-way component analysis results, and procedures to decide on the numbers of components to use. This culminated in a paper (Kiers & van Mechelen, 2001) where three-way analysis in all its steps is explained to a non-technical audience of psychologists.

Current research focuses on development and application of multiway analysis techniques in special contexts (e.g., neuroimaging), and evaluate them in terms of their usability. Special attention will be given to the assessment of the reliability of their outcomes, using bootstrap procedures or cross-validation.