摘要
本文介绍了实验室能力验证常用的三种稳健统计方法——四分位法、迭代算法A和Q/Hampel方法。通过实际检测数据计算示例,比较了3种稳健统计方法的计算结果。对于对称、单峰分布,四分位法和算法A可以较好估计均值和标准差,但对于大比例(>20%)离群值,或者有小众数的情形Q/Hampel方法则更为适用。
The paper introduces three popular robust statistical methods, including quantile method, algorithm A and Q/Hampel method. Using the three methods two examples are presented to compare the results of different calculations. For the dataset following symmetric and unimodal distribution, quantile method and algorithm A have good estimation of mean and standard deviation, but for the datasets with high proportion of outliers(>20%) or minor modes Q/hampel method is more applicable.
引文
[1]GB/T 6379.5-2004,测量方法与结果的准确度(正确度与精密度)第5部分:确定标准测量方法精密度的可替代方法.
[2]GB/T 28043-2011,利用实验室间比对进行能力验证的统计方法.
[3]ISO 13528:2015 Statistical methods for use in proficiency testing by interlaboratory comparison.
[4]国家认证认可监督管理委员会编著.国家认监委实验室能力验证技术报告汇编[M].中国质检出版社,中国标准出版社,2014.