Use of Trimean in Theil-Sen Regression Analysis

Abstract

Theil-Sen regression analysis is the most preferred method in non-parametric regression analysis. In the Theil-Senmethod,calculationsaremadewiththemedianparameter. In this study, it was proposed to calculate the trimean parameter instead of the median parameter. In this way, the effects of the outliers in the data on the model are fully reflected. In applications of one real-life and two simulation data, the results obtained with the use of trimean were more successful. It is recommended to use the trimean parameter instead of the median parameter in data structures with an excess of outliers.