This part of GB/T6379 provides a basic method for estimating the bias of a measurement method and the laboratory bias when applying a measurement method. GB/T 6379.4-2006 Accuracy (Trueness and Precision) of Measurement Methods and Results Part 4: Basic Methods for Determining the Trueness of Standard Measurement Methods GB/T6379.4-2006 Standard Download Decompression Password: www.bzxz.net
This part of GB/T6379 provides a basic method for estimating the bias of a measurement method and the laboratory bias when applying a measurement method.

ICS 03.120.30
National Standard of the People's Republic of China
GB/T6379.4--2006/ISO5725-4:1994 Partially replaces GB/T6379-1986
GB/T11792-—1989
Accuracy (trueness and precision) of measurement methods and results-Part 4:Basic methods for the determination of the truenessof a standard measurement method（ISO5725-4:1994.IDT）
Promulgated on November 13, 2006
General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China Administration of Standardization of the People's Republic of China
Implementation on April 1, 2007
GB/T 6379.4—2006/1SO 5725-4:1994Foreword
Normative references
Determination of the bias of a standard measurement method based on interlaboratory trials4.1
Statistical model
Requirements for reference materials
Considerations in the design of experiments for estimating the bias of a measurement methodCross-reference with GB/T 6379.1 and GB/T 6379.2Required number of laboratories
Statistical evaluation
Interpretation of the results of the statistical evaluation
Determination of the bias of a standard measurement method for a single laboratory5.1
Implementation of the experiment:
Cross-reference with GB/T 6379.1 and GB/T 6379.2 Number of cross-referenced test results
5.4 Selection of standard materials
5.5 Statistical analysis
6 Report to the leadership team and the decisions made by the leadership team Report of the statistical expert
6.2 Decisions taken by the leadership team
Application of trueness data
Symbols and abbreviations used in GB/T 6379
Appendix A (Normative Appendix)
Appendix B (Informative Appendix)
B.1 Description of the test
B.2 Evaluation of precision
B.3 Evaluation of trueness
B.4 Progress One-step analysis
Example of accuracy test
Appendix C (informative appendix)
Derivation of formula
C.1 Formula (5) and (6) (see 4.5)
C.2 Formula (19) and (20) (see 5.3)Appendix D (informative appendix)References
GB/T6379.4—2006/1SO5725-4:1994GB/T6379 "Accuracy (Trueness and Precision) of Measurement Methods and Results" is divided into the following parts. Its structure and corresponding international standards are as follows:
Part 1: General principles and definitions (ISO572 5-1:1994, IDT); Part 2: Basic methods for determining the repeatability and reproducibility of standard measurement methods (ISO5725-2:1994, IDT); Part 3: Intermediate measures of the precision of standard measurement methods (ISO5725-3:1994, IDT); Part 4: Basic methods for determining the trueness of standard measurement methods (ISO5725-4:1994, IDT); Part 5: Alternative methods for determining the precision of standard measurement methods (ISO5725-5:1998, IDT); Part 6: Practical application of accuracy values ​​(ISO5725-6:1994, IDT). This part is Part 4 of GB/T6379.
This part of GB/T6379 is equivalent to the international standard ISO5725-4:1994 "Accuracy of measurement methods and results (trueness and precision)-Part 4: Basic methods for determining the trueness of standard measurement methods". Part 1 to Part 6 of GB/T6379 replace GB/T6379-1986 and GB/T11792-1989 as a whole. In the standard, the original precision is expanded to add correctness, collectively referred to as accuracy; in addition to the repeatability condition and reproducibility condition, the intermediate precision condition is added.
Appendix A of this part is a normative appendix, and Appendix B, Appendix C and Appendix D are informative appendices. This part is proposed and coordinated by the National Technical Committee for Standardization of Statistical Methods. Drafting units of this part: Institute of Mathematics and Systems Science, Chinese Academy of Sciences, China National Institute of Standardization, Guangdong Entry-Exit Inspection and Quarantine Bureau. The main drafters of this part: Feng Shiyong, Ding Wenxing, Yu Zhenfan, Jiang Jian, Xiao Hui, Chen Yuzhong, Li Chengming. This part was first published in 2006.
GB/T6379.4—2006/ISO5725-4:1994 Quote
0.1GB/T6379 uses the two terms "trueness" and "precision" to describe the accuracy of a measurement method. Trueness refers to the degree of agreement between the (arithmetic) mean of a large number of test results and the true value or accepted reference value; while precision refers to the degree of agreement between test results. 0.2GB/T6379.1 gives general considerations for the above quantities, which will not be repeated in this part of GB/T6379. GB/T6379.1 should be read in conjunction with all other parts of GB/T6379 (including this part) because GB/T6379.1 gives basic definitions and general principles. 0.3 When the true value of the measured characteristic is known or can be inferred, the trueness of the measurement method is of concern. Although for some measurement methods, the true value may not be known exactly, it is possible to know an accepted reference value of the measured characteristic. For example, if an acceptance reference value can be determined using appropriate standard materials (reference substances/standard materials) or by reference to another measurement method or by preparing a known sample. By comparing the acceptance reference value with the level of results given by the measurement method, the correctness of the measurement method can be assessed. Correctness is usually expressed in terms of bias. For example, in chemical analysis, if the measurement method used cannot completely extract a certain element, or if the presence of one element interferes with the determination of another element, bias will occur. 0.4 This part of GB/T 6379 considers the following two measures of correctness: a) Bias of the measurement method: Where the measurement method may be biased, regardless of when and where the measurement is made, attention should be paid to the "bias of the measurement method" (as defined in GB/T 6379.1). For this purpose, experiments involving multiple laboratories are required, and GB/T 6379.2 has more explanations on this.
Laboratory bias: Measurements made by a single laboratory can reveal "laboratory bias" (as defined in GB/T 6379.1). If the laboratory bias is estimated based on -b)
tests, it should be noted that this estimate is only valid for the time the test is carried out. If it is to be demonstrated that the laboratory bias does not change, further formal testing is required, and GB/T6379.6 describes the relevant methods. I
1 Scope
GB/T6379.42006/ISO5725-4:1994 Accuracy (Trueness and Precision) of Measurement Methods and Results Part 4: Basic Methods for Determining the Trueness of Standard Measurement Methods
1.1 This part of GB/T6379 provides basic methods for estimating the bias of a measurement method and the laboratory bias when applying a measurement method.
1.2 The measurement methods involved refer specifically to measurement methods that measure continuous quantities and take only one measurement value as the test result each time, although this value may be the result of a group of observations. 1.3 In order to facilitate the measurement of the same conditions, the measurement method is standardized and all measurements are performed according to the standard method. 1.4 Bias is a certain estimate of the ability of a measurement to give a correct (true) value. When reporting the test results and bias of a measurement method according to a measurement method, it means that the measurements were made using exactly the same method on the same characteristic. 1.5 The provisions of GB/T 6379 apply only to situations where the reference value can be used as an agreed true value. For example, a) a certified standard (material) with known characteristics or c) a material (substance/material) whose characteristics are measured according to a measurement method or a known sample prepared according to a reference measurement method: or a measurement method to measure 1.6 This part of GB/T 6379 only considers materials where the measurement at a certain time is affected by another known bias. The situation where the bias at a certain level is calculated is not applicable to the situation where the level of one characteristic is affected by another characteristic. GB/T 6379.6 gives Note 1: In GB/T 637.2, the normative reference document is used (i.e., the interaction effect is not considered). The comparison of the two measurement methods in this part only considers the bias at a certain time, so the number of the relevant level is omitted. The clauses in the following documents become the clauses of this part through reference in this part of 6379. For all the referenced documents with dates, all subsequent amendments (excluding outdated content) or revisions are not applicable to this part. However, parties to agreements based on this part are encouraged to study whether the latest version of these documents can be used. For all undated references, the latest version applies to this part.
GB/T 3358.1-1993 Statistical terminology Part 1: General statistical terminology GB/T 6379. 12004
(ISO5725-1:1994, IDT)
Accuracy (trueness and precision) of measurement methods and results Part 1: General principles and definitions
GB/T 6379.2-2004 Accuracy (trueness and precision) of measurement methods and results Part 2: Basic methods for determining the repeatability and reproducibility of standard measurement methods (ISO5725-2:1994.1DT) 3 Definitions
The definitions given in GB/T 3358.1 and GB/T 6379.1 are still applicable in this part of GB/T 6379. The symbols used in GB/T 6379 are given in Appendix A. GB/T 6379.4—2006/ISO5725-4:19944 Determination of the bias of a standard measurement method based on interlaboratory trials 4.1 Statistical model
In the basic model described in 5.1 of GB/T 6379.1, the total mean m can be expressed as: mo
Where:
—Accepted reference value of the measured characteristic:
Bias of the measurement method.
Thus the model is rewritten as:
y=u+a+B+e
Formula (2) is used for the case of concern. Where B is the laboratory component of the bias, that is, the component in the test result that represents the interlaboratory variation. The laboratory bias 4 is given by formula (3):
So the model can be recorded as:
ymu+a+e
Formula (4) is used for the case of concern 4.
4.2 Requirements for standard materials
When standard materials are required, the conditions of 4.2.1 and 4.2.2 shall be met. Standard materials shall be homogeneous. 4.2.1 Selection of standard materials
4.2.1.1 Standard materials shall have known characteristic values ​​(e.g., concentration, content) at each level within the range of levels to which the standard measurement method is to be applied. In some cases, it is important to use a set of standard materials in the evaluation test, each corresponding to a different level of the characteristic, because the bias of the standard measurement method at different levels may be different. The matrix of the standard material should be as close as possible to the matrix of the material being measured by the standard measurement method, such as carbon in coal and carbon in steel. 4.2.1.2 A sufficient number of standard materials shall be prepared for the entire test, and a certain amount of surplus shall be left in the sieve to cope with unexpected needs. 4.2.1.3 Wherever possible, the properties of the reference material should remain as stable as possible throughout the test. There are three situations: The property is stable: no precautions need to be specified in advance: a) The demonstrated value of the property may change due to storage conditions: the container should be stored in the manner described in the instructions before and after opening: e) The value of the property changes at a known rate: a statement is required along with the reference value to determine the value of the property at a specific time. 4.2.1.4 Any possible difference between the specified value and the true value of the property is expressed in terms of the uncertainty of the reference material (see ISO Guide 35). This is not considered in the method given here. 4.2.2 Inspection and distribution of reference materials Before distribution, the reference materials need to be reduced. Special care should be taken at this time to avoid introducing any additional errors. Reference should be made to relevant national (international) standards for sample reduction. The distribution of sample units should be randomly selected. If the measurement process is non-destructive, it is possible to send the same sample of standard material to each laboratory participating in the laboratory test, but this will extend the time period of the entire test.
4.3 Considerations in test design when estimating the bias of the measurement method 4.3.1 The purpose of the test is to estimate the amount of bias of the measurement method and determine whether it is statistically significant. If it is statistically significant, a further goal is to determine the maximum amount of bias that cannot be detected with a certain probability based on the experimental results. 4.3.2 The test arrangement is almost exactly the same as the precision test described in 5.1 of GB/T6379.2, with the only difference being: u) an additional acceptance reference value is required:
b) the number of laboratories participating in the test and the number of test results should meet the requirements in 4.5. 2
4.4 Cross-reference with GB/T6379.1 and GB/T6379.2 GB/T6379.4-2006/IS05725-4:1994 Chapter 6 of GB/T6379.1-2004 and Chapters 5 and 6 of GB/T6379.2-2004 apply to this part. In this case, "precision" and "repeatability and reproducibility" in the text of the above two standards should be replaced by "accuracy". 4.5 Number of laboratories required
The number of laboratories required and the number of test results required at each level are related to each other. 6.3 of GB/T6379.1-2004 discusses the number of laboratories required. The following is a guide to determining the number of laboratories. Based on the test results, the minimum number of laboratories required to detect a predetermined amount of bias with a high probability is The number of force and test results should satisfy the following relationship (see Appendix C):
where:
the predetermined bias that the experimenter hopes to detect from the test results; the reproducibility standard deviation of the measurement method:
A——a function of and, given by formula (6):
where:
(ncy-D+]
Y=an/o,
Table 1 gives the value of A.
represents the estimated value of the measurement method bias. The uncertainty value A = 2
(6)
...(7)
For the value of. determined in advance by the experimenter, the ideal situation is that the number of laboratories and the number of repeated tests in each laboratory satisfy or (5). However, for practical reasons, the number of laboratories is usually selected based on the available resources and the need to reduce it to one. A compromise between two satisfactory levels. If the reproducibility of the measurement method is poor, it is unrealistic to require a high degree of confidence when estimating the bias. In most cases, is greater than, (that is, γ is greater than 1). At this time, the number of tests per laboratory at each level ~ greater than 2 will not be significantly improved compared to = 2. 4.6 Statistical evaluation
The test results should be handled in the manner described in GB/T6379.2. In particular, when outliers are detected, all necessary steps should be taken to check the causes of their occurrence. At the same time, the appropriateness of the accepted reference value adopted should be reassessed. 4.7 Interpretation of statistical evaluation results
4.7.1 Precision test
The precision of the measurement method is represented by s (estimated value of repeatability standard deviation) and (estimated value of reproducibility standard deviation). In formulas (8) to (10), it is assumed that each The number of tests in the laboratories is equal. If not, the corresponding formulas given in GB/T 6379.2 are used to calculate s and 5R.
4.7.1.1 The repeatability variance estimate for each laboratory is calculated according to the following formula3
GB/T 6379.4--2006/ISO5725-41994Ya
where and are the variance and mean of the test results obtained by the first laboratory, respectively. m (8)
(10)
For the variance, the Cochran test described in GB/T 6379.2 is used to check whether there are significant differences in the variances within the laboratories. At the same time, the h and post plots of Mamda described in GB/T 6379.2 are used to conduct a more comprehensive check on potential outliers.
If the repeatability standard of the standard measurement method cannot be determined in advance according to the method of GB/T6379.2, S, shall be used as its best estimate.2 method to determine, can be used to calculate the ratio: calculated value. If the repeatability standard deviation of the standard measurement method C = s / a
boundary value is compared:
the test statistic C
where. () is the freedom degree
a if C ≤ c
b) C
in the previous case
reasons to investigate
4. 7. 1. 2 force parameter
is not significant
4+（11)
distribution of a point to make the number. Unless otherwise specified, the reproducibility standard deviation will be used to assess the bias of the measurement method. In the latter case, it is necessary to estimate the reproducibility variance of the laboratory that produces the fluctuations before further investigation. If the reproducibility standard deviation of the standard measurement method cannot be obtained, the best estimate of G is calculated. The reproducibility of the standard quantitative method can be evaluated indirectly by calculating the following:
Compare the test statistic C with the following critical value.-(12)
...(13)
If the test statistic C is determined in advance by the method of T6379.2, it can be considered as the standard deviation of the other, and determined by the method of GE/T6379.2-(1-1/n)s
G-(11/n)c,
X-)()/y
a quantile. Unless otherwise stated, a is assumed to be taken as the distribution of x-() with degrees of freedom = 0.05.
a) If C≤Cm, then s-(1-1/n)s is not significantly greater than -(1-1/n)ab)) If C>Cen, then s-(1-1/n)s is significantly greater than -(1-1/m). In the former case, the repeatability standard deviation, together with the reproducibility standard deviation of 0, will be used to assess the trueness of the measurement method: In the latter case, the working conditions of each laboratory must be carefully checked before the bias of the measurement method can be assessed. There may be cases where some laboratories do not use the required equipment or do not work under the prescribed conditions. In chemical analysis, problems may arise, for example, from improper control of temperature, humidity or contamination. As a result, the test must be repeated to obtain the desired precision value. 4.7.2 Standard measurement method bias estimate Www.bzxZ.net
The estimate of the bias of the measurement method is given by the following formula: 4
where
can be positive or negative.
GB/T6379.4-2006/ISO5725-4:1994-15)
If the absolute value of the bias estimate is less than or equal to half the length of the uncertainty interval (as defined in ISO Guide 35), the bias is not significant.
The variation of the estimated value of the bias of the measurement method comes from the variation of the results of the measurement process, and its size is expressed by its standard deviation. When the precision value is known, its calculation formula is:
When the precision value is unknown, the calculation formula is (ok-(1-1/n)d)
The appropriate 0095% confidence interval is:
The
of the bias of the measurement method is unknown, so its estimated value
is used, where A is given by formula (6).
includes the measurement method
If the confidence interval includes
instead of the
value, it is necessary to use a 8/sr
significant, then the odd deviation is significant. The bias of the standard measurement method is not greater than that of the standard measurement method at a confidence level of -5%. Determination of the bias of a laboratory The inter-laboratory precision test of GB/T 6379.2 is as follows. The laboratory bias is the same as that of the laboratory. The test can be used under different conditions. 5.1 Implementation of the test The repeatability of the standard measurement method should be strictly checked under repeatability conditions. This also includes comparison (between different laboratories) Before estimating the accuracy, the actual indoor standards used by the paired laboratories should be compared with those used by the standard measurement method. The only substantial difference between the repeatability standard deviation of the referenced measurement method and the precision test in GB/T 6379.2 is the need for the laboratory to make the same measurements. Except for the measurement of a single fact, it is not worthwhile to put too much precision on the test. Perhaps more attention should be paid to the longevity of GB/T 6379. The laboratory bias is not practical. 5.2 Cross-reference with GB/T 6379.1 and GB 6379.2 When the repeatability is poor in reference to GB/T 6379.1, the "sticky" in the text should be changed to GB/T 6379.2. "Accuracy\or repeatability and reproducibility" is replaced by "accuracy\. Since the number of laboratories in GB/T6379.2 is equal to 1 at this time, the two roles of "executive person in charge" and "measurement person in charge" can be performed by one person.
5.3 Number of test results
The uncertainty of the laboratory bias estimate depends on the repeatability of the measurement method and the number of test results obtained. In order for the test results to detect a predetermined bias with a high probability (see Appendix C), the number of test results n should satisfy the following relationship: Awo,1.84
Where:
the predetermined laboratory bias that the experimenter hopes to detect from the test results, o
the repeatability standard deviation of the measurement method, and
（20）
GB/T 6379.4—2006/IS05725-4:19945.41 Selection of standard materials
When standard materials are used, the requirements of 4.2.1 also apply. 5.5 Statistical analysis
5.5.1 Test of within-laboratory standard deviation
For n test results, calculate the mean yw and the estimated value sw of the within-laboratory standard deviation aw: w
—yw)
(22)
The outliers in the test results should be carefully checked according to the Grubbs test described in 7.3.4 of G13/T6379, 2-2004
If the repeatability standard deviation a of the measurement method is known, the estimated value sw can be evaluated by the following method: Calculate the ratio
C= (sw/a,)2
Compare it with the critical value
Ce. - xa- (s/v
-（23)
, where -() is the 1-α quantile of the x distribution with a degree of freedom of = one. Unless otherwise stated, α is assumed to be 0.05.
a) If cm, then sw is not significantly greater than o
b If CCm. then sw is significantly greater than a,
In the former case, the standard deviation of the repeatability of the measurement method will be used to assess the laboratory bias: in the latter case, repeated tests should be considered to verify that all steps in the standard measurement method are implemented normally. 5.5.2 Estimation of laboratory bias
The estimate of laboratory bias 4 is given by the following formula: Ayw-p
The variation of the estimate of laboratory bias comes from the variation of the results of the measurement process and can be expressed by its standard deviation. When the repeatability standard deviation is known, the calculation formula is:
5g = 8. / Vn
When the repeatability standard deviation is unknown, the calculation formula is: w/yn
The 95% confidence interval of laboratory bias can be calculated as: Z-Awa,+Awa
where Aw is given by formula (20). If 6, is unknown, its estimated value 5. is used instead. If the confidence interval contains 0, the laboratory bias is not significant at the confidence level of 5%, otherwise the bias is significant. GB/T6379.6 further considers laboratory bias. 6 Report to the leadership group and decisions made by the leadership group 6.1 Report of the statistical expert
After completing the statistical analysis, the statistical expert should submit a report to the leadership group. The report should include the following: 1) A full description of the opinions on the benchmarking method learned from the operator and/or the person in charge of the measurement; b) A full description of the outlier laboratories and the excluded laboratories; + (25)
.. (26)
(27)
A full description of each outlier value and/or statistical outlier found, and whether they have been explained, corrected or 6
eliminated:
Final result table including mean and precision measurement: GB/T6379.4—2006/ISO5725-4;1994Statement on whether the bias of the standard measurement method relative to the adopted acceptance reference value is significantIf the bias is significant, an estimate of the bias should be reported for each level.
6.2 Decision taken by the leadership team
The leadership team shall discuss the report of the statistician and make a decision on the following issuesAre the test results consistent?If there is a clear inconsistency, is it due to an inappropriate description of the standard measurement method? a)
What action should be taken for the excluded outlier laboratory?Do the test results of the outlier laboratory and/or the opinions of the operator and the person responsible for implementation indicate that the standard measurement method needs to be improved? If so, what should be improved?Do the results of the accuracy test confirm that the measurement method is acceptable as a standard measurement method? What actions should be taken before publication? Application of accuracy data
According to the requirements of Chapter 7 of GB/T6379.1-2004 Appendix A of GB/T6379.4-2006/ISO5725-4:1994
(Normative Appendix)
Symbols and abbreviations used in GB/T6379
The distance in the relationship s=u+bm
A is used to calculate the uncertainty coefficient b of the estimated value. The slope in the American system s=a+bm
B represents the deviation component of a laboratory test result from the total mean value (biased laboratory Component) B, represents the component of B when all factors remain unchanged under the condition of intermediate precision) B) Ba) +... represents the component of B when the factor changes under the condition of intermediate precision. The distance in the relationship Igs=c+dlgm
C, C, C test statistic
Ca.CmC is the critical value of the test
CD of the critical value of the probability P
CR, the limit of the probability P
d The rate
in the relationship igs
m occurs in each test result. The critical range coefficient of the random error class
G Grubbs
Mandelbrot's quantiles of the distribution
statistic
Consistency test statistics
Mandelbrot's internal consistency test statistic
Lower control limit (action limit or warning limit)
Grand mean of the test characteristic: level
in the middle
number of interactions
number of factors considered by each laboratory in the
analysis (i.e., the test results on
units) Number of results
Number of laboratories participating in the interlaboratory trial
Number of levels of quality tested in the interlaboratory trialRepeatability limit
Reproducibility limit
RMEstimated value of standard deviation of standard material (reference substance/standard material)
Predicted value of standard deviation
TTotal
Number of test targets or groups
UJCI, upper control limit (action limit or warning limit)Weights in the calibration regression
Range of a set of test results
Data used for Grubbs test
Test results
Tip: This standard content only shows part of the intercepted content of the complete standard. If you need the complete standard, please go to the top to download the complete standard document for free.