Comparison of different robust estimators on performance evaluation results of drinking water proficiency testing in Indonesia: A case study

• Main Authors: Elishian, Christine, Ridwan, Yohanes Susanto
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Case studies; Drinking water; Estimators; Outliers (statistics); Performance evaluation; Robustness; Statistical analysis; Statistical methods; Uncertainty; Visual observation
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0108853

Proficiency testing (PT) is considered as a strategic instrument for the successful implementation of national quality system in many countries, including Indonesia. Since initiated at the end of 1990, various PT activities have been carried out. Among various assigned values that could be used in PT, consensus value is commonly used in Indonesia. In the first period (end of 1990-2011), statistical methods commonly used to get the consensus value were methods with pre-elimination of specific values as outliers, followed by a simple statistic. Some of them are based on visual observation to make the data selection, prior to performance evaluation using z-score. After SNI ISO/IEC 17043:2010 was introduced, robust statistic estimators have been introduced to PT providers, such as Algorithm A, Q/Hampel, etc. However, those estimators were not yet many implemented on Indonesian PT providers, mostly because of unaccustomed. This study aimed to compare the use of recent robust estimators with the results obtained from the conventional one. Data from PT in drinking water in 2011 is taken as a case study. The study revealed that the statistic method used in the original work resulting in a higher number of outliers compared to the results obtained from the Alg-A and Q/Hampel method. Uncertainty evaluation of the assigned value in the original work was also evaluated, and it was showed that uncertainties of assigned values for all parameters were met the criteria as defined in SNI ISO 13528:2016