Evaluation of the Statistic Parameter of Rationality in Proficiency Testing by the Empirical Model

Proficiency testing is a good approach to find the difference between individual laboratories in the same field. However, it was found that statistical results in the proficiency testing reports organized by CNAS were distorted since some laboratories reported the analysis data not based on their own experiments. How to identify this situation and ensure all participants to obtain a fair evaluation is a problem that needs to be urgently solved for CNAS. In order to evaluate the rationality of the statistical results, a ratio (be called as H) of target standard deviation (robust standard deviation, s_R) to those calculated by the empirical model modified by experimental data collected from a large number of proficiency testing reports in the domestic environmental field as a criterion is proposed in this paper. Thirteen sets of data in recent years were used to validate this criterion. The results show that target standard deviations of ten sets of data are within the normal range and the part of robust parameters in three sets of data are much smaller compared to those calculated by this model. The causes for these situations were analyzed. It was suggested that the target standard deviation be calculated based on algorithm recommended by ISO 13528, and the Z-score be calculated by the calculated value from the model instead of the target standard deviation from experimental data when H is smaller than 0.5. The validation results showed that it is possible to evaluate the rationality of a statistical parameter in proficiency testing based on a modified Horwitz function, effectively identifying and preventing the situation in which the statistical results are distorted by some anthropogenic factors.