An outlier is an observation unusually far from the behavior expected under a statistical model. A large regression residual can identify a candidate, but data validation, regression leverage, and the possibility of model misspecification should be examined before excluding it. An outlier need not be an influential observation.
New to topics? Read the docs here!