Robust Outliers Detection Method for Skewed Distribution

Authors

  • Prem Junsawang Department of Statistics, Faculty of Science, Khon Kaen University, Khon Kaen, Thailand
  • Mintra Promwongsa Department of Statistics, Faculty of Science, Khon Kaen University, Khon Kaen, Thailand
  • Wuttichai Srisodaphol Department of Statistics, Faculty of Science, Khon Kaen University, Khon Kaen, Thailand

Keywords:

Robust outlier detection, skewed data, split interquartile range

Abstract

The aim of this study is to propose the robust outliers detection method called MH boxplot for skewed distribution. The proposed method is modified from Hubert’s boxplot by embedding the Bowley coefficient, the ratio of lower split interquartile range and upper split interquartile range into the fences of the boxplot. The performance of the boxplot is evaluated by the percentage of outlier ratio mean in three cases of simulated data (truncated, uncontaminated and contaminated data) and real data. Furthermore, the existing boxplots for outliers detection are used to make a comparison with the MH boxplot as well. The results from simulated and real data show that the MH boxplot efficiently detects outliers and is robust to skewness of data over the other boxplots for any sample
size. Moreover, the MH boxplot efficiently detects outliers as the shape of real data.

Downloads

Download data is not yet available.

Downloads

Published

2021-06-29

Issue

Section

Articles