Automatic detection of discordant outliers via the Ueda's method

Abstract

The importance of identifying outliers in a data set is well known. Although variousoutlier detection methods have been proposed in order to enable reliable inferencesregarding a data set, a simple but less known method has been proposed by Ueda(1996/2009). Since this new method, called Uedas method, has not been systematicallyanalysed in previous research, a simulation study addressing its performance androbustness is presented. Although the method was derived assuming that theunderlying data is normally distributed, its performance was analysed using data fromvarious outlier-prone distributions commonly found in several research fields. Theresults obtained enable us to define the strengths and weaknesses of the methodalong with its limits of applicability. Furthermore, an unforeseen field of application ofthe method, which requires further studies was also identified.