农残水平测试中数据多态性分析及指定值估计研究  

Study on Multimodality and Assigned Value Establishment in Pesticide Residues Proficiency Test Scheme

作　　者：纪雷[1] 王建华[1] 刘学惠[1] 林雨霏[1] 昃向君[1] 杜恒清[1] 刘心同[1] 张艺兵[1] 梁成珠[1] 何桂华[1] 

机构地区：[1]山东出入境检验检疫局技术中心,山东青岛266002

出　　处：《分析测试学报》2007年第4期503-507,共5页Journal of Instrumental Analysis

摘　　要：对全国农残水平测试中毒死蜱、氯氰菊酯数据进行统计分析,在数据统计分布特征研究基础上,使用内核密度估计进行数据多态性分析,使用bootstrap模拟取样法对数据样本值重复取样,以获得稳健的水平测试样品待测物含量代表值估计、标准误差及置信区间描述,证明以bootstrap模拟取样法获取的均值与标准偏差作为有限单次样本代表值是合理、有效的,解决了四分位稳健统计方法对非正态多态分布代表值估计不稳定问题及取样理论中取样样本数限制的瓶颈,为能力验证计划指定值的获取提供了一种新方法。Data profiles of chlorpyrifos and cypermethrin collected from a nation-wide pesticide resi- dues proficiency test schema was investigated by statistical techniques. Based on the statistical distribu- tion of data, a kernel density estimation, coupled with bootstrap resampling method was introduced to acquire the assigned values of chlorpyrifos and cypermethrin content in distributed sample matrix. By means of its resampling simulation mechanism, the bootstrap method resolved the problems arisen from the muhimodality of data distribution and limitation of sampling. This method has the prior advantage of giving a robust descriptions to explain the central tendency and the variation of data profiles. A bootstrap mean and interval were provided and discussed in detail.

关 键 词：农残 水平测试 指定值 多态性 内核密度估计 Bootstrap模拟取样法 

分 类 号：TB99[一般工业技术—计量学] S482[机械工程—测试计量技术及仪器]