油料发射光谱仪数据比对的方法研究  

Study on the Distributions of Interlaboratory Comparisons Data of Oil Atomic Emission Spectrometer

作　　者：刘云涛[1,2] 田洪祥[1] 董庆山 

机构地区：[1]海军工程大学船舶与动力学院,湖北武汉430033 [2]海军91959部队装备部,海南三亚572016

出　　处：《计量学报》2012年第4期364-367,共4页Acta Metrologica Sinica

基　　金：湖北省自然科学基金（2010CDB01505）

摘　　要：在油料发射光谱仪的比对研究中，发现部分实验数据不服从正态分布，不能直接运用基于正态分布假设的比对方法，如z比分数检验法和h值检验法进行比对。基于高等统计学中的污染分布模型和极大似然估计原理，对不服从正态分布数据的总体分布估计模型进行了修正。即，假设数据服从含两个或多个正态分布的污染分布，且应以污染分布中系数最大的正态分布作为对总体分布的估计。卡方检验结果表明，4组不服从正态分布的元素浓度数据，有3组服从污染分布。It ＇ s found that part of the data didn＇ t fit normal distribution in the process of interlaboratory comparisons of oil atomic emission spectroscopies, so that, methods which are based on normal distribution hypothesis, such as Z-test and h- test,couldn＇t be used to the data. Based on polluted distribution class model in statistics and the principle of maximum likelihood estimation,the distribution model which based on normal distribution hypothesis were mended. It＇s assumed that the non-normal distribution data should fit polluted distributions contained 2 or more normal distributions. And the normal distribution, whose coefficient is the biggest in the polluted distribution, should be considered as the estimated distribution of population. Chi-square test results showed that, of the 4 non-normal distribution elements, there are 3 elements fitted polluted distribution, and the other one didn＇ t.

关 键 词：计量学 油料发射光谱仪 实验室比对 污染分布 卡方检验 

分 类 号：TB99[一般工业技术—计量学]