On a new approach to the analysis of variance for experiments with orthogonal block structure. III. Experiments in row-column designs