Remarks on the test of significance for the method of paired comparisons