有效自由度在评定LDH催化活性不确定度中的应用

The application of effective degrees of freedom in evaluating measurement uncertainty of catalytic activity of LDH

目的探讨在乳酸脱氢酶(LDH)催化活性测量不确定度评定中是否有必要计算有效自由度。方法首先测量LDH有证参考物质计算效能函数(En),对本室建立的IFCC参考方法进行确认,然后用参考方法对同一份人混合血清样本中LDH催化活性水平重复测量20次,计算A类自由度(vA)、A类相对不确定度[uA(rel)]、B类相对合成不确定度[ucB(rel)]及有效自由度(veff),另外再计算uA(rel)与ucB(rel)不同比值下,不同vA对应的veff及95%置信水平下的包含因子(kP)。结果 En小于1,方法学性能符合要求;LDH催化活性水平vA、uA(rel)、ucB(rel)和veff分别为19、0.55%、1.04%和398,此时随机变量服从正态分布,在95%置信水平下,kP为1.96;当uA(rel)与ucB(rel)比值大于1/2时,随着vA的增大,kP逐渐缩小,当vA足够大时(≥100),kP为1.96,当vA较小时,kP明显大于1.96,且随着uA(rel)与ucB(rel)比值的增加而扩大;当uA(rel)≤1/2ucB(rel),kP为1.96。结论当vA≥100或uA(rel)≤1/2ucB(rel)时,在LDH催化活性水平测量不确定度评定中无需计算veff;当uA(rel)与ucB(rel)比值大于1/2且vA较小时,应首先计算veff,再查阅t界值表得veff对应的kP值后计算其实际测量不确定度值。

Objective To investigate whether calculating effective degrees of freedom is necessary for evaluating the measurement uncertainty of LDH catalytic activity concentration. Methods Firstly, En was calculated to verify the IFCC reference method by using LDH certificate reference material. Secondly, the catalytic activity concentrations of LDH in human mixed serum was repeatedly measured for 20 times with the reference method to calculate the degree of freedom vA of type A uncertainty, type A relative uncertainty UA（rel）, type B relative combined uncertainty ucR（rel） and effective degrees of freedom veff. In addition, the veff and coverage factor k （95% confidence level） under different ratios of UA（rel） and UcB（rel） were calculated. Results When En was less than 1, the performance of the reference method should be excellent, and the values of vA, uA〈（rel）, and UcB〈（rel） were 19, 0.55%, 1.07% and 398, rela- tively, while therandom variables followed normal distribution. Because veff was more than 100, kp was equal to 1.96 under 95% confidence level. When UA（rel） was more than half of UcB（rd） , k. gradually reduced with the increasing of vA. k, was equal to 1.96 if vA was large enough （≥ 100） , but the value of kp was obviously larger than 1.96 and gradually enlarges with the increase of the ratio of uA〈（rel） and UcB〈（rel） if vA was relative small. When UA（rel） was less than or equal to half of ucB（rel） , k, was also equal to 1.96. Conclusion When vA was equal to or large than 100 or uA（rel） was less than or equal to half of UcB〈（rel） , there was no need to calculate veff. When uA（rel） was large than half of ucB（rel） and vA was relative small, and k, under 95% confidence level, the veff should be calculated before evaluating measurement uncertainty.