Comparison of statistical methods for outlier detection in proficiency testing data analysis of fineness in ordinary Portland cement

The study evaluates the performance of participating laboratories in a Proficiency Testing Program for Ordinary Portland Cement (OPC) fineness analysis using various statistical methods. The study found that nIQR, which considers both within- and between-laboratory variance, is most suitable for outlier detection in the dataset. The analysis showed that different statistical approaches can significantly affect the evaluation of laboratory performance, emphasizing the importance of selecting an appropriate method for proficiency assessment in OPC fineness testing. The study compared Cochran’s, Grubb’s, Hampel’s, Dixon’s, Mandel’s, nIQR, robust algorithm-A, and classical z-score methods for outlier detection and performance assessment. The z-score indicates laboratory performance relative to others, with values between ±2 and ±3 considered questionable and those outside ±3 considered unsatisfactory. The study highlights the importance of an effective evaluation system and the significance of selecting the appropriate statistical method for proficiency assessment.