A trimmed mean is the arithmetic mean computed after discarding specified proportions of the smallest and largest observations.
If are the order statistics, the
-trimmed sample mean is
for . Symmetric trimming uses the same
at both ends; asymmetric trimmed means
are also possible.
Trimming reduces sensitivity to extreme observations at the cost of discarding data. The ordinary sample mean is obtained when ,
while increasingly heavy symmetric trimming moves the statistic toward the sample
statistical median.