Outliers in Biometrical Data - Two Real Examples of Analysis

Abstract

The problem of finding and identifying outliers in multivariate measurement data is as old as the data analysis itself. The question: "what is an outlier" has many answers, is is especially difficult when considering high-dimensional data. In the paper a search for outliers in two real data sets is shown. It is stressed that identifying outliers should not be done on the basis of asymptotical cutoffs derived under assumption of normality of the analyzed data.