If you have some distribution, the standard deviation is defined as symmetric about the mean; what measure do people use for different positive and negative error values?