摘要
对ISO13528和澳大利亚国家检测机构协会(NATA)所用的能力验证尧敦图的统计模型进行了讨论,分别利用霍特林T2检验和二维正态分布推导出尧敦椭圆方程的公式,使用Excel软件绘制出尧敦椭圆,参与能力验证的各个实验室的分析数据也在同一个图中显示出来。利用坐标变换,还得到了其他不同坐标下的尧敦图。在一个尧敦图中同时绘制出置信水平分别为95%和99%的尧敦椭圆,直观显示出合格数据、存疑数据和不合格数据。ISO 13528中,ZA,B是系统误差和随机误差的组合,利用主成分分析,可以将他们区分和表示出来。由于原始数据不总是正态分析,且采用了不同的数学模型,NATA方法得到的尧敦椭圆和ISO13528得到的尧敦椭圆有时是不同的。
The statistical model of Youden plot for proficiency testing adopted by ISO13528 and the National Association of Testing Authorities, Australia （NATA） were discussed. The elliptic equa- tions of Youden plot were separately deduced by Hotelling＇s T2 test and bivariate normal distribution. Youden ellipse was drawn by Excel software, and the analytical data from all laboratories participating in the proficiency testing was shown in the same plot. Meanwhile, Youden plots with different coordi- nates were obtained by coordinate transformation. In the same Youden plot, the Youden ellipses with confidence levels of 95% and 99% were drawn out respectively to give a visual display of qualified da- ta, questionable data and failure data. In ISO 13528, ZA,B was the combination of systematic error and random error, which could be distinguished and represented with principal component analysis. Since the original data was not always normal distribution and the mathematical models are different, the Youden ellipses obtained by NATA method were sometimes different from those obtained by ISO 13528.
出处
《冶金分析》
CAS
CSCD
北大核心
2013年第12期74-81,共8页
Metallurgical Analysis
关键词
能力验证
尧敦图
霍特林T2检验
主成分分析
二维正态分布
proficiency testing
Youden plot
Hotelling＇s T2 test
principal component analysis
biva-riate normal distribution
