Outlier detection in circular regression model using minimum spanning tree method

Abstract

The existence of outliers in circular-circular regression model can lead to many errors, for example in inferences and parameter estimations. Therefore, this study aims to develop new algorithms that can detect outliers by using minimum spanning tree method. The proposed method is examined via simulation study with different number of sample sizes and level of contaminations. Then, the performance of the proposed method was measured using "success" probability, masking effect, and swamping effect. The results revealed that the proposed method were performed well and able to detect all the outliers planted in various conditions.