Outlier Removal

Outliers should only be removed when there is a strong belief that they do not come from the true data-generating process. Such a case may arise, for example, due to measurement errors or other errors in the data collection.

As a result, robustness to outliers may or may not be a good thing.