TREBIG: A 360/75 FORTRAN program for three-mode factor analysts designed for big data sets

In the analysis of multifactor, longitudinal data, it has long been the practice of investigators to factor analyze the data in each time period separately and then compare the factor structures by eye. This has obvious problems and various measures of similarity were devised, none of which really worked properly. In recent years, other procedures have been proposed to better handle these sorts of data (i.e., Corballis, 1970; Jöreskog, 1970; Tucker, 1966). Tucker’s solution is termed “3-mode factor analysis,” which is essentially the Eckart-Young decomposition of a set of matrices that are derived by collapsing a tensor of order three. It has the advantage of combining the factor analyzes of each time period and indicating how the changes occur.