Ⅰ型极小值分布样本异常数据的检验

Detection of Outliers from the Sample of Type I Minimum-Value Distribution

针对Ⅰ型极小值分布样本的多个异常数据,提出了一种新的检验方法.首先寻找到总体参数的具有较好稳健性的估计量,然后在此基础上构造出检验统计量,进一步求出了该检验统计量精确的概率密度函数和大样本情形下的近似分布.由于检验统计量中的核心统计量——样本分位数,对于异常数据的干扰具有一定的抵抗力,因此利用该方法可以达到有效的检验效果.

The method for the detection of many outliers from the sample of typeⅠminimum-value distribution is put under discussion. A new testing method is presented：first,the estimator with a relatively satisfactory robustness was determined for the population parameter; second,on the basis of the first step,test statistics was constructed and then the accurate density function and the approximate distribution of the said test statistics were worked out. Since the sample fractile,the core stasistics in the test statistics,has resistance immunity against the disturbance of the outliers,such a method proves to have high testing efficiency.