consider the mad statistic. what would happen if you did not take absolute values before calculating the sum of differences between group means? why would this not be a useful calculation without taking absolute values?