The Boxplot: A Robust Exploratory Data Analysis Tool for the Definition of the Threshold for Outlier Data

Abstract

Data from chemical analyses are subject to variability due to sampling, sample preparation, analytical method, variation in the phenomena under study, etc. In classical statistics, prior to the analysis of data, it is required to transform it in order to satisfy the Gaussian distribution assumption. However, this is not always posible owing to the presence of "wild" values, a phenomenon omnipresent in studies such as in environmental pollution and many other specialised fields. This note introduces an unconventional technique for the analysis and interpretation of univariate data.