Construction and Applications of Stairboxplot for Exploratory Data Analysis

Abstract

The classical construction of boxplot requires estimates of five robust statistics of interest namely, the first quartile, the median, the third quartile, the lower fence and the upper fence. The fence estimate is usually dependent on the three quartiles and is utilized to visually identify outliers in a batch of univariate dataset. Some scholars are critical of the limitation of boxplox to display individual data points, density of observations and distributional shape in multiple batch comparison among to mention. In this paper display enhancement to address the limitations of classical boxplot is proposed according to a new construction method called stairboxplot. The construction begins with display of four stairs of consecutive boxes according to quadbins to replace box and whiskers in the classical boxplot construction and an inscription of individual observations using scale adjusted outlyingness estimate of each data point. The advantage of stairboxplot as a data display toolkit was explored using simulation and real life dataset.