Outlier

Outliers lie far from the rest of the data. In a list of monthly sales figures between 40,000 and 60,000, a month with sales of 250,000 would be an outlier.

Outliers can arise for several reasons. They may be genuine but unusual events, such as a one-off large order or a market crash. They may be errors, such as a typing mistake that adds an extra zero. Or they may come from a different group than the rest of the data, such as a corporate customer included among individual customers.

Outliers can strongly affect some statistical measures. The mean, standard deviation, range and correlation coefficient are all sensitive to extreme values, while the median is much less affected. A single outlier can make a regression line misleading.

Dealing with outliers requires judgment. Errors should be corrected. Genuine outliers should not simply be deleted, because they may contain important information. In finance, rare extreme events, such as large market falls, are exactly what risk managers most need to understand. Analysts often check results with and without outliers to see how much they influence conclusions.