Detecting multiple outliers in linear regression using a cluster method combined with graphical visualization (Q2271696): Difference between revisions

A new algorithm is proposed for multiple outliers detection in regression models. The idea is to apply a hierarchical clustering algorithm to the scatterplot of standardized predicted values and residuals (obtained by ordinary least squares) in order to find a single largest cluster which is considered as a set of inliers. Then the \(t\)-statistic is used to test all other observations as candidate outliers individually against the group of inliers. Results of real data sets analyses are presented.