The Application Research Of Statistical Method In Laboratory Proficiency Testing

Proficiency Testing(PT)is a project to judge the abilities of laboratories and inspection bodies by using the comparison information between each other,but also is one of the necessary conditions for the accreditation bodies to join the international Mutual Recognition Agreement(MRA).Proficiency Testing is organized and implemented by the China National Accreditation Board for Laboratories(CNAL).In the ability validation activities,in order to evaluate the ability of the laboratory more accurately,we need to make adequate analysis of the data,so it is important to make full use of the statistical method.In proficiency testing,most institutions use most is the traditional robust statistical methods,namely the Z value.This method is simple,easy to grasp.However,as the method is based on the assumption that the data follows the normal distribution,not all the data follow the normal distribution in practice.Meanwhile,there are differences in the abilities to verify the final number of participating labora-tories and recycling cycle.Therefore,part of the problems still occurs in the course of practice by using the traditional robust statistical method,which leads to that the organizers make non-objective results in the use of profici-ency testing and unrealistic understanding of the participating laboratory's testing capabilities.Given this,it is very vital to find a more effective statis-tical method.At present,we have made some statistical research methods,such as principal component analysis method,non negative matrix factorization method,kernel density estimation method and Bootstrap method.But we do not know whether these methods are more appropriate or other defects,this requires us to further discussion.At present,there is no document on the systematic test and summary of these methods.This paper mainly discusses these four methods,these four methods are applied in the data processing of proficiency testing,and comparison and analysis,Identify the strengths and weaknesses of the four methods.Eventually make a general analysis and summary,in order to establish the ability to apply different statistical met-hods to verify the results of distribution.This thesis is divided into four chapters.The first chapter mainly intro-duces the ability to verify the relevant knowledge,research and development,as well as the specific methods and research content of this paper;The second chapter mainly introduces the traditional robust methods,principal component analysis method,non negative matrix factorization method,kernel density estimation method and Bootstrap method principle;In the third chapter mainly introduces the application of traditional robust statistical methods and other four statistical methods in proficiency testing activities by data validation;The fourth chapter summarizes the paper,proposes systematic analysis and summary,so as to establish the statistical analysis method which is applicable to the distribution of different capability valid-ation results.