用Horwitz经验模型评估环境监测分析方法标准再现性标准偏差数据合理性

再现性标准偏差(SR)是分析方法基本性能参数之一,构成方法标准中再现性限基础.对应于不同浓度的再现性限是日常质控规范重要依据之一.以环境监测领域土壤,沉积物及固废样品中无机元素不同分析方法标准为例,用Horwitz经验模型(Ⅱ类)和H判据(实验统计值SR与模型计算值之比),考察了已颁布执行的标准文本和正在生态环境部官网...

Evaluation for rationality of the reproducibility standard deviation in environmental monitoring analysis methods standard with Horwitz empirical model

The standard deviation of reproducibility (s_R) was one of the basic performance parameters of the analytical method, and it was the basis to calculate the reproducibility limit in the method standard. The reproducibility limit of different concentration was one of important basis for daily quality control specifications. Taking the standard texts for different analysis methods of inorganic elements in soil, sediment and solid waste samples in the field of environmental monitoring as an example, the Horwitz's empirical model (II) and H criterion (the ratio of experimental statistical value s_R to calculated value of the model) were used to examine the rationality of the reproducibility standard deviation in the standard text that had been published and implemented, and method validation report for public comment and advice published in the official website of the Ministry of Ecology and Environment.It was showed that H>4 and H<0.5 were common phenomena. When H>4, it implied: (1) there might be "outliers" in the data involved in statistical calculation; (2) there might be "double peaks" in the data, and s_R data had deviated from the normal range. When H<0.5, it implied: whether the precision of the experimental data was too optimistic, and whether there were data that should not be deleted. The outlier in the data could be judged by combining the daily quality control specifications and the fact that whether the H value was in the normal range. The raw measurement data from the method verification report was used as examples to explain the identification process of outliers. In order to avoid the interference of various human factors those directly leaded to lower s_R, the actual (soil, sediment and solid waste) samples should be used as much as possible in precision verification experiments. When certified reference materials was selected, it was not required that the reported results must be located in the uncertainty interval of the certified value of certified reference materials, and it was considered as meeting the requirements for satisfying the daily quality control specifications. Using password coding was also an effective way to effectively avoid human interference. The Horwitz empirical model and H criterion were useful tools for quickly checking the rationality of reproducibility standard deviation data.