稳健统计与格拉布斯准则在能力验证结果分析中的应用

Application of Robust Statistics and Grubbs Criterion in Proficiency Testing Results Analysis

分别运用稳健统计(四分位和迭代)法、格拉布斯准则剔除离群值后用经典统计法对热轧带肋钢筋的6组数据进行了统计分析。结果表明:数据组中Z比分值大于5的数据,为离群值。数据组总体服从正态分布或接近正态分布的,3种统计方法结果基本吻合;不存在离群数据的情况下,建议采用稳健统计(四分位或迭代)法进行统计;存在离群数据的情况下,应采用格拉布斯准则剔除离群值后用经典统计法进行统计。对于数据明显偏离正态分布的,不存在离群数据的情况下,建议采用迭代稳健统计法进行统计;存在离群数据的情况下,应采用格拉布斯准则剔除离群值后用经典统计法进行统计。

The data of the 6 groups of hot rolled ribbed bar were analyzed respectively using robust statistics of four quantile, robust statistics with iterative method, classical statistical method. The results show that the data of Z in the data set is larger than 5, which is outlier, The data are subject to or close to the normal distribution, the results of the three statistical methods are basically consistent ; if there is no outlier data, it is recommended to use robust statistical method （four quantile method or iterative method） for statistical analysis; If there is outlier data, the Grubbs criterion should be used to eliminate the outliers and the classical statistical method should be used for statistics; The data significantly deviate from the normal distribution, if there is no outlier data, it is recommended to use iterative method for statistical analysis; If there is outlier data, the Grubbs criterion should be used to eliminate the outliers and the classical statistical method should be used for statistics.