A Comparison of the recent algorithms for the identification of outliers in data

ABSTRACT Ph.D. THESIS A COMPARISON OF THE RECENT ALGORITHMS FOR THE IDENTIFICATION OF OUTLIERS IN DATA GÜLSEN KIRAL DEPARTMENT OF MATHEMATICS INSTITUTE OF NATURAL AND APPLIED SCIENCES ÇUKUROVA UNIVERSITY Supervisor: Assoc. Prof. Dr. Nedret BİLLOR Year: 2003, Pages: 155 Jury: Assoc. Prof. Dr. Nedret BİLLOR Prof. Dr. Fikri AKDENİZ Prof. Dr. H. Altan ÇABUK Prof. Dr. Müjgan TEZ Prof. Dr. Refik BURGUT In this thesis we examine outlier detection methods in multivariate and regression data and present an extensive literature on them. Since there has been a fast growing interest in the outlier detection methods for regression data for ten years we mainly focus on the outlier detection methods in regression data in this thesis. We conduct an extensive simulation study to assess the performances of the multiple outlier detection methods for regression data, that are either most recently published or most frequently cited in the literature. Furthermore in the context of multivariate data we propose a new outlier detection algorithm in principal component analysis called BACON robust principle component analysis. We also carry out a simulation study to assess the performance of the proposed method and use some data sets to evaluate the applicability of the method. Key Words: influential observations, detection of outliers, multivariate data, principle component analysis, regression. IV