Comparative Analysis of three outlier detection methods in univariate data sets

In the field of geomatics engineering and signal processing, observational data can bring a lot of information. However, the abnormal data (i.e. outliers) may be acquired due to human carelessness or limiting condition, causing interference to subsequent research. To solve this problem, outlier detection methods have been proposed to detect and remove outliers. Three common outlier detection methods (i.e. Z-score method, boxplot method and median absolute deviation method) in signal processing are introduced in this paper. A comparison of the three methods is conducted through experimental evaluation with two sets of experiments. The results of experiments show that the number of outliers detected with median absolute deviation method significantly outperform those of Z-score and boxplot methods. It shows that the median absolute deviation method can more effectively detect and remove outliers so that the more reliable and accuracy results can be obtained.