Alternating least squares estimation for the single domain in DEDICOM model

DEDICOM (Harshman, 1978, 1981b) is a generalization of multidimensional scaling capable of analyzing asymmetric relationships among a set of things (e.g., the number of telephone calls between cities, confusions among stimuli, strength of implications among propositions, or preferences among products). However, existing estimation procedures for DEDICOM have several limitations: most of the procedures are only approximately-least-squares, and none of them are well suited for ignoring the (often problematical) diagonal cells of the input data matrix.