robust estimator
an estimator for an unknown characteristic in a population that is less likely than others to be influenced by violations of assumptions about the sample data on which it is based. For example, rather than using a mean to describe central tendency in a distribution with outliers, a researcher might choose a median, a trimmed mean, or a Winsorized mean. Each option reflects a different way of handling the outliers, such as downweighting them, replacing them with a new value, or ignoring them. See also resistant estimator.