On some clinically useful measures of the accuracy of diagnostic tests