Dr. Aylin KOLBAŞI1, Prof. Dr. Aydın ÜNSAL2
1.Turkish Statistical Office, Ankara/Türkiye.
2.Hacı Bayram Veli Unıversity, Ankara/Türkiye.
Abstract: In many data analysis tasks a large number of variables are being recorded or sampled. One of the first steps in obtaining a consistent analysis is to identify observations that are far from the center. Outlier values often carry important information, even if they are considered to be errors or problems. However detected outliers lead to incorrect modeling, biased parameter estimates, and incorrect results. For this reason, it is very important to identify them before modeling and analyzing. In this study it has been tryied to determine the method of detecting outliers which can detect the outliers in foreign trade data in the most accurate way. For this, z-score, median z-score, box-plot, adjusted box-plot, B-H method, k-means clustering method and robust regression methods were applied to the foreign trade data set and the application results of outlier detection methods were compared by considering some criteria. In the light of these criteria, the most appropriate outlier detection methods that can be applied to the foreign trade data set have been determined.
Keywords: Outlier, foreign trade, outlier detection methods.
Pages: 213 – 234 | Full PDF Paper