It combines maximum likelihood estimation of the parameters of matrix and 3rd-order tensor normal distributions with unstructured factor variance-covariance matrices, two procedures, and unbiased modified likelihood ratio testing of simple and double separability for variance-covariance structures, two procedures.