Font Size: 

An approach for clustering three-way data is discussed. The approach, which is based on mixtures of matrix-variate distributions, uses an iterative subset log-likelihood approach to detect and trim outliers.