摘要
探讨了非参数统计方法在白砂糖色值测定能力验证中的应用价值。通过实际制备样品并依据CNAS-GL003指南,分别采用参数方法(F检验、t检验)与非参数方法(Kruskal-Wallis H检验、Mann-Whitney U检验)对样品的均匀性和稳定性进行统计评估。结果表明,在稳定性分析中2类方法结论一致,但在均匀性检验中非参数方法识别出参数方法未检出的变异,提示样品存在不均匀性。进一步将样品间变异纳入Z比分数修正后,发现传统Z值与修正值z'之间存在5%~9%的偏差,表明显著影响能力评价准确性。研究表明,非参数方法对非连续、偏态或含离群值的数据具有更强稳健性,可作为参数方法的重要补充,提高能力验证统计结果的科学性与可靠性。
This study investigates the value of nonparametric statistical methods in proficiency testing for the color value determination of white granulated sugar.Following the preparation of actual samples and in accordance with the CNAS-GL003 guidelines,both parametric methods(F-test,t-test)and nonparametric methods(Kruskal-Wallis H test,Mann-Whitney U test)were employed to statistically evaluate sample homogeneity and stability.The results demonstrated that both methodological approaches yielded consistent conclusions regarding stability analysis.However,in homogeneity testing,the nonparametric method detected variations that were not identified by the parametric methods,indicating potential inhomogeneity in the samples.Furthermore,after incorporating inter-sample variation into a modified Z-score,a deviation of 5%~9%was observed between the traditional Z-values and the modified values(z′),demonstrating a significant impact on the accuracy of proficiency evaluation.The study confirms that nonparametric methods offer greater robustness for discontinuous,skewed,or outlier-containing data and can serve as a vital complement to parametric methods,thereby enhancing the scientific rigor and reliability of statistical results in proficiency testing.
作者
邓倩南
陈其钊
肖爱玲
王桂华
李家威
DENG Qiannan;CHEN Qizhao;XIAO Ailing;WANG Guihua;LI Jiawei(Institute of Biological and Medical Engineering,Guangdong Academy of Sciences,Guangzhou,510316,Guangdong,China;Engineering Technology Research Center of Sugarcane Industry China National Light Industry Council,Guangzhou 510316,Guangdong,China)
出处
《甘蔗糖业》
2025年第6期78-83,共6页
Sugarcane and Canesugar
基金
食糖标准样品研发与质量控制技术研究(0525270026)。
关键词
白砂糖
色值
非参数统计
能力验证
White granulated sugar
Color value
Nonparametric statistics
Proficiency testing
