Sampling as a source of measurement uncertainty: techniques for quantification and comparison with analytical sources
Sampling as a source of measurement uncertainty: techniques for quantification and comparison with analytical sources
复制标题
采样作为测量不确定度的来源：量化技术以及与分析源的比较
DOI：
--
复制
发表时间：
1998
期刊：
影响因子：
--
通讯作者：
M. Ramsey
中科院分区：
文献类型：
--
作者：
M. Ramsey
一个教程回顾了目前的方法估计测量不确定度由一次抽样。回顾和解释了当前用于描述不确定度和分析数据质量的术语。本文详细介绍了一种估计抽样不确定度的基本方法，并给出了一个应用于测试数据集的实例。这种方法采用了一式两份的部分样品，并对这些样品进行了进一步的化学分析。稳健方差分析(ANOVA)用于估计总的测量不确定度，并量化初级采样和化学分析过程中产生的不确定度的贡献。单因素方差分析程序和测试数据以电子方式提供，以便能够应用该方法。讨论了这种基本方法的假设和局限性，包括它不能估计抽样偏差。讨论了更复杂的方法，包括估计采样和分析中的系统误差对不确定度的贡献。从抽样理论和地统计学两方面对抽样不确定度估计的其他方法与这些方法进行了比较。对作为两个不确定度来源的采样和化学分析进行了比较，它们之间是相对的，也是相对于测量的总体方差的。给出并讨论了测量方差占总方差比例的理想最大值和最小值，以及抽样方差和分析方差的相对贡献率的合适性准则。
A tutorial review is presented of current methods for the estimation of measurement uncertainty due to primary sampling. Current terminology used in the description of uncertainty and analytical data quality is reviewed and explained. One basic method for the estimation of uncertainty in sampling is described in detail with a worked example of its application to a test dataset. This method employs the taking of a proportion of samples in duplicate, with the further duplication of chemical analysis on these samples. Robust analysis of variance (ANOVA) is applied to estimate the total measurement uncertainty and also to quantify the contributions to that uncertainty which arise from the processes of primary sampling and chemical analysis. The ANOVA program and test data are available electronically to enable application of the methodology. The assumptions and limitations of this basic method are discussed, including its inability to estimate sampling bias. More sophisticated methods are discussed that include the estimation of the contributions to uncertainty from systematic errors in both sampling and analysis. Other approaches to the estimation of uncertainty from sampling, from both sampling theory and geostatistics are compared with these methods. The comparison is made between sampling and chemical analysis as the two sources of uncertainty, relative to each other, and relative to the overall variance of the measurements. Fitness-for-purpose criteria are given and discussed for the ideal maximum and minimum values of the proportion of the measurement variance to the total variance, and the relative contributions of the sampling and analytical variances.