A Method for Detecting Outliers in Survey Data by Ratio Tests

A new method based on ratio tests is proposed. This paper discusses the basic principle and basis of the method, explores the mechanism, condition, and the way to improve its detection effectiveness. It has been shown that when a certain condition is satisfied the ratio indicator will have lower coefficient of variation so that outliers can more easily be detected. Reasonably selecting ratio indicators is the key to effectively detect outliers. The proposed method is also suitable for data with probability distribution unknown. Empirical study shows that the method is simple, practical, easy to use, and has application value.