This is the process of eliminating outliers. In fact this is essential before using statistical calculation because all inferential (deduction) is false if your datas don't eliminate exceptional cases: statistics are about common cases not exception. But statistic can be used to detect outliers. There are more or less complex procedure. For most simplest procedure datas above 3 standard deviations should be considered as outliers. Also 2 consecutive datas above two standard deviations should also be suspicious. This presuppose that the mother population is normal. If it is not the case transformation like log-normal should be applied before. If it is difficult to normalise at least it is possible to make groups and whatever the law th mean of a group even as small as 4 should follow rather well the normal law.