Identifying regression outliers and mixtures graphically

Regressions in practice can include outliers and other unknown subpopulatio n structures. For example, mixtures of regressions occur if there is an omi tted categorical predictor, like gender or location, and different regressi ons occur within each category. The theory of regression graphics based on central subspaces can be used to construct graphical solutions to long-stan ding problems of this type. Tt is argued that in practice the central Subsp ace automatically expands to incorporate outliers and regression mixtures. Thus methods of estimating the central subspace can be used to identify the se phenomena, without specifying a model. Examples illustrating the power o f the theory are presented.