Several Methods For Testing Outliers

Outlier testing is to test some outliers whether are in a group of samples or not.Outlier is the individual value of datas and its value is obvious away from the others. Theexistence of outliers decrease the quality of data to a great extend, and makes thecorresponding results of data analysis significantly change more, which will lead topeople’s incorrect conclusions. Thus, it’s the primary work in statistics analysis.This paper investigates the basic theories of hypothesis testing method for outliers inthe observational data. Concerning several important distributional, such as UniformDistribution, Extreme Value Distribution, Two-Parameter Exponential Distribution. Thetesting method is discussed if a outlier occur in the sample observational data. Several newpreferable test statistics based on the order statistic or the superior estimate of theparameter are given. Its main results are as follows:Firstly, this paper discussed the procedures of testing outliers from the UniformDistribution, on the basis of order statistics, a new test statistic is given. And we get thestatistics’ precise distributions by the property of the order statistic. And then, we comparethe efficiency between the order statistic and the Dixon statistic. The result is that theorder statistic has a better efficiency.Secondly, this paper discussed the procedures of testing outliers from the ExtremeValue Distribution, mainly for the Dixon test statistic is proposed to improve the higherefficacy of the test statistic, derived the exact distribution of the test statistic about theExtreme Value Distribution improved, to discuss the efficacy of two statistics contrast,statistics show that improved more power-efficient.Lastly, this paper discussed the testing outliers from the Two-Parameter ExponentialDistribution, identify population parameter has better robustness of the estimated amount,constructed based on the new test statistic, derived the precise density function wasestablished under the original assumptions, and provide a basis for assuming the test.