Proficiency testing for determination of pesticides residues in black tea: Comparison of three robust statistical approaches based on ISO 13528 to estimate the consensus values for small number of participants

Abstract

In Indonesia, proficiency testing (PT) provider for determination of pesticides residues in tea is still limited. In addition, many proficiency testing schemes still use consensus value from participants’ results as the assigned value rather than reference assigned value due to some technical and economical issues. However, this mode of assigned value may lack of reliability when the number of participants are small and the test results are highly varied. In these conditions, many PT providers were not able to perform the statistical data evaluation causing the needed PT became unavailable. This study is aimed to investigate the suitable statistical approach in estimating the consensus value for small number of participants in PT for determination of pesticides residues in black tea. Three robust statistical approaches recommended by ISO 13528: variants of Algorithm A, Hampel method, and M estimator were compared to evaluate the participants’ data from the PT schemes organized by Research Center of Chemistry-LIPI in 2014 and 2015. All of the robust statistical approaches showed to be able to assess the performance of participants. The Hampel method and M estimator were the most adequate robust statistical approaches in obtaining the closest consensus values compared to the homogeneity test results, stability monitoring results, and the assigned valued in the real PT scheme, for endosulfan sulphate and bifenthrin, respectively.