Relevance of Outlying and Influential Observations for Regression Analysis

Abstract

As the name suggests, the outlying observations (commonly called “outliers”) are the ones which “lie out” from the remaining observations in the sample. In regression models, the outliers are observations with large residuals (Makridakis et al. 1998). First, we will discuss the nature of the outliers occurring in the case of the single variables, which we will call the univariate outliers. Then we will proceed to discussing the more subtle type of outliers, which occur in the case of multivariate analyses.