Robust Determination of Mean Trimmed Mean

Next: Likelihood Based Robust Up: 5: Goodness of Previous: Robust Determination of

One method that is more robust than the arithmetic mean, is the trimmed mean
where pN events are left off each end.
For the special case of distributions symmetric about their mean, the trimmed will also have mean and error .
The variance of will be larger than for Gaussianly distributed .
But as you investigate distributions broader than Gaussians you will find the variance of becoming smaller than that of .
A special case of is the median (i.e., middle value of when they are numerically ordered) gotten with .