The Application Of Principal Component Analysis In Proficiency Testing

Proficiency testing program is a kind of implementation plan for evaluating laboratories test capability by inter-laboratory comparison. At present, proficiency testing activities are widely launched in our country. They are used to supervise and evaluate the test capacity and the test level of laboratory by departments and institutions at different levels on the one hand, and regarded as a reliable and objective technique to control the quality of laboratories on the other hand. The development and utilization of proficiency testing program provide powerful technical support for the metrological analysis and evaluation of laboratories in our country.In the methods of implementing proficiency testing program, the robust statistical analytical method is the most common one, for it can be used to compare quantitative experimental data respectively, with a speedy, simple and direct feature. However, when robust statistical methods are applied in the proficiency testing field, many problems happened in this process cannot be avoided, such as incapacity to describe experimental data dependency, lack of overall judgment and evaluation.At present, some domestic scholars use principal component to do comprehensive assessment of proficiency testing, but all of them do it on the condition of analyzing and dealing with a particular case with fairly complete data. Whereas the processed data is mass, the result of applying principal component analysis is usually inaccurate, part of which is rather different from that of applying traditional robust statistical analysis.In order to resolve the problems mentioned above, this thesis, based on principal component analysis, takes the leading in introducing the concepts like "degree of satisfaction","satisfaction index" as well as provides readers with computing methods of LPTSI and LPTSR. Furthermore, by using analyzing drawing software——SAS and MATLAB, this thesis has found new distribution regularities of principal component score scatter plot in proficiency testing field. After describing the lab code score distribution of different satisfaction indexes, this thesis has also given a new reliable proficiency testing assessment method.On the basis of robust statistical analysis, LPTSI and LPTSR, which have combined with a multivariate statistics method——principal component analysis, not only reveal the direct feature of robust statistical analysis, but also possess the advantages of principal component analysis. Meanwhile, this new technique has integrated with the concept of "degree of satisfaction" in psychology, which makes the analysis result further close to the actual situation.