Detecting Outliers With The Least Trimmed Squares

Abstract

Classical outlier tests made classically based on the least-squares (LS) have significant disadvantages in some situations. The results of adjustment computation and classical outlier tests performed with classical methods are deteriorated when observations are not distributed independently and this distribution is not normal. To detect outliers that do not have a normal distribution, the robust techniques that are not sensitive to outliers have been developed. The least trimmed squares (LTS) known as having a high-breakdown point have been dealt with in this study. Adjustment computation has been carried out based on the least-squares (LS) and the least trimmed squares (LTS). A certain polynomial with arbitrary values has been used. In this way, the performances of these techniques have been investigated.