deviantoutlier
Deviantoutlier is a term used in data analysis to refer to data points that are extreme in their value and simultaneously deviate from expected patterns according to a given model. The term is not part of formal statistical nomenclature but is used in some communities to stress the dual nature of such observations.
In practice, a deviating outlier might have two attributes: a univariate extreme value and a large residual
Detection approaches combine univariate outlier tests with regression diagnostics or residual analysis. For multivariate settings, one
Applications occur across fields such as finance, engineering, and healthcare, where unusual observations may indicate fraud,
Limitations include dependence on model choice and threshold settings, which can render the designation unstable. Misclassification