The Research Of Outlier Detections Using Uniformly Distribution

Outlier detection is to test some outliers whether are in a group of samples or not. Outlier is the exceptional datum in a passel of data and its value is obvious away from the others. In sampling, Outliers are mixed into the normal samples possibly caused by inadvertence or contrived reason etc. So, it is important and basal to test whether there are outliers in a group of samples in statistics analysis, and the correctness of further analysis can be ensured when excluding outliers.In this paper, some common used outlier detections are classified by the distribution which the outlier detection is appropriate for , and each of outlier detections is discussed individually; furthermore, a new outlier detection — the detection using uniformly distribution is discussed, which includes the detection using superior and inferior α/2 position point and the detection using order statistics. When the population distribution is given without unknown parameters, samples can be tested whether have outliers, moreover, whether have the upper outliers and the lower outliers by the detection using uniformly distribution. Therein, the test using superior and inferior α/2 position point can avoid the effect of Masking and Swamping.The efficiency of one test can be estimated by testing lots of experimental data. So, a great deal of uncontaminated and contaminated samples are tested whether have outliers by the outlier detections introduced in this paper, and the samples obey the common distribution. Then, each of outlier detections is further analyzed according to the given results.