发布时间 : 星期一 文章EP5-A2-翻译稿-定量测量方法的精密度性能评价-end更新完毕开始阅读861517e2524de518964b7dca
数据计算表＃1：精密度评价实验 每日一批
分析物/浓度：葡萄糖/高值 工作日 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 总的WR变异（S2r）＝7.225 总的WR标准差（Sr）＝2.69 日均数的变异（B2）＝8.88 日均数的标准差（B）＝2.98 批内的变异 8.0 0.5 32.0 32.0 8.0 0.5 12.5 0.0 8.0 2.0 0.5 0.5 2.0 4.5 0.5 2.0 4.5 0.0 2.0 24.5 批内的均数 244 242.5 243 245 244 244.5 243.5 245 241 245 251.5 248.5 241 247.5 247.5 239 242.5 244 240 243.5 总均数=244.13 仪器：XYZ
数据计算表＃2：精密度评价实验 每日一批
分析物/浓度：葡萄糖/高值
仪器或实验室精密度标准差的计算： S2r（源自表1）：7.225 B2（源自表1）：8.88
N（＃每次运行的重复次数）：2
仪器：XYZ
ST?B2??3.53N?121Sr?8.882?(7.225) N2ST自由度的计算：
I＝20
ME＝S2r＝7.225
MD＝N?B2＝17.76
I????N?1??ME?MD??
T＝I?MD2?N?1??ME2?I?1＝
2(7.225?17.76)2?7.225?202??17.76?192
＝
624.25
19.21＝32.49 T为32。
NCCLS consensus procedures include an appeals process that is described in detail in Section 8 of the Administrative Procedures. For further information contact the Executive Offices or visit our website at www.nccls.org. Summary of Consensus/Delegate Comments and Committee Responses
EP5-A2: Evaluation of Precision Performance of Quantitative Measurement Methods; Approved Guideline—Second Edition
General
1. A glossary would be useful either as a ―lead-in‖ section or with definitions clearly called out in the text.
? A section on “Definitions” has been added as recommended.
2. The protocols for evaluated precision in both EP5-A and EP5-A2 do not strictly apply to hematology testing, since it is not possible to have the same blood specimen tested for several times within a day for 20 days. We can have protocols to estimate repeatability (as defined in EP5-A2), but most of the time other variance estimates (day-to-day, laboratory-to-laboratory, etc) will be confounded.
? This concern is addressed in the third sentence in the Scope, which states: “These procedures may not be appropriate for some quantitative methods for which adequate test materials do not exist.”
3. I would have preferred to see procedures for the estimation of measurements of uncertainty (ISO).
? The committee agrees that MU (measurements of uncertainty) is an important concept and that EP5-A2 is closely related. This is addressed in the fifth paragraph of the Foreword, which states that the precision estimates from EP5-A2 are components of measurement uncertainty, but that GUM (ISO Guide to the Expression of Uncertainty in Measurement) estimates of MU that comply with ISO procedures are beyond the scope of the document. This is because laboratory-specific estimates of uncertainty may contain components of error other than precision, and may involve corrections for bias.
4. The recommended number of days (20) should be given in terms of ?repeats for a certain factor‘ (or degrees of freedom) to make it more general to any deviation from the base protocol. In addition, a factor that is expected to introduce more variability should be repeated more times than some other that is fairly constant.
? See response to comment number 6, below. See also Appendix C for discussion of modifications.
5. Formulas for calculating precision apply to experiments conducted exactly as described in the protocol. These formulas are totally or partially useless and sometimes misleading when even slight deviations from the protocol (e.g., having three replicates instead of two) are implemented. Currently, there are faster and more elegant ways for obtaining estimates of variance components. I think discussions should be based on the use of commercial statistical software (similar to EP6-A), while these formulas can be put together in an appendix for reference.
? The formulae for unbalanced designs are quite complex and beyond the scope of this document. To be fully general for all situations would reduce the usefulness for less statistically sophisticated users, as explained in the Foreword (third paragraph). While many statistical software packages correctly calculate variance components, not all of them do. Further, these software packages can be difficult to interpret and are best left to the attention of professional statisticians.
Section 7, Statistical Power of Precision Estimates
6. Section 7 does not provide any ?concrete‘ guideline for sample size. The discussion is purely academic.
? It is the intent of the committee to specify a minimum number and guidance for when a situation might require larger numbers. Goal-based experimental design is beyond the scope of the document, but might be considered in a future version.
Section 7.2, Statistical Comparison with the Manufacturer
7. Make degrees of freedom consistent. Use either 100 or 40 for both.
? Section 7.2 presents guidance that is consistent with the general protocol, which allows the user