INTERNATIONAL ISO STANDARD 5725-4 Second edition 2020-03, Accuracy (trueness and precision) of measurement methods and results — Part 4: Basic methods for the determination of the trueness of a standard measurement method Exactitude (justesse et fidélité) des résultats et méthodes de mesure — Partie 4: Méthodes de base pour la détermination de la justesse d'une méthode de mesure normalisée Reference number ISO 5725-4:2020(E) © 180 2020ISO 5725-4:2020(E) A COPYRIGHT PROTECTED DOCUMENT © 1802020 All rights reserved. Unless otherwise specified, or required in the context of its implementation, no part ofthis publication may be reproduced or utilized otherwise in any form or by any means, electronic or mechanical, including photocopying, or posting fon the internet or an intranet, without prior written permission, Permission can be requested from either 1SO at the address below or ISO's member body in the country ofthe requester. ISO copyright office CP 401 » Ch. de Blandonnet 8 (CH-1214 Vernier, Geneva Phone: #4122749 01 11 Fax: +41 22 749 09.47 Email: copyright@[Link] Website: [Link], Published in Switzerland it © 150 2020 ~All rights reservedISO 5725-4:2020(E) Contents Foreword Introduction. 1 Scope 2 Normative references 3 Terms and definitions. 4 ‘Symbols 5 Determination of the bias of a standard measurement method by an interlaboratory experiment 5.1 Experimental design considerations. 5.1.1 Objective. 5.1.2. Layout of the experiment. 5.1.3. Cross-references to ISO 5725-1 and ISO 5725-2. 5.2 The statistical model. 5.3 Required number of laboratories and measurements 5.4 Requirements of the accepted reference value. 5.4.1 Approaches to assigning the accepted reference value 5.4.2 Materials used in the experiment. 5.4.3 Requirements of measurement uncertainty of the accepted reference value. 55 Carrying out the experiment, 5.5.1 Evaluation of precision 5.5.2 Check of precision. 5.5.3. Estimation of the bias of the standard measurement method, 5.5.4 Example Determination of the laboratory bias of one laboratory using a standard measurement method. 6.1 Experimental design considerations. 6.1.1 Objective. 6.1.2 Layout of the experiment. 6.1.3 Cross-references to ISO 5725-1 and ISO 5725-2 62 Thestatistical model 63 Number of measurement results 64 — Requirements of the accepted reference values. 65 Carrying out the experiment, 65.1 Check of the within-laboratory standard deviation. 6.5.2. Estimation of the laboratory bias Report to the panel and decisions to be taken by the panel 71 — Cross-reference to ISO 5725-2 7.2 Report by the statistical expert. 7.3 Decisions by the panel Annex A (informative) Derivation of formulae Annex B (informative) Example of an accuracy experiment. Bibliography @ 180 2020 ~ Al rights reserved Page z NNR SSanarid sa area ewww 10 10 10 iL 11 i 12 12 13 14 14 14 15 18 26 itiISO 5725-4:2020(E) Foreword ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies (ISO member bodies). The work of preparing International Standards is normally carried out through ISO technical committees, Each member body interested in a subject for which a technical committee has been established has the right to be represented on that committee. International organizations, governmental and non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization. The procedures used to develop this document and those intended for its further maintenance are described in the ISO/IEC Directives, Part 1. In particular, the different approval criteria needed for the different types of ISO documents should be noted. This document was drafted in accordance with the editorial rules of the ISO/IEC Directives, Part 2 (see [Link]/directives). Attention is drawn to the possibility that some of the elements of this document may be the subject of patent rights. ISO shall not be held responsible for identifying any or all such patent rights. Details of any patent rights identified during the development of the document will be in the Introduction and/or on the ISO list of patent declarations received (see [Link]/patents). Any trade name used in this document is information given for the convenience of users and does not constitute an endorsement, For an explanation of the voluntary nature of standards, the meaning of ISO specific terms and expressions related to conformity assessment, as well as information about ISO's adherence to the World Trade Organization (WTO) principles in the Technical Barriers to Trade (TBT) , see wwwniso -org/iso/[Link]. ‘This document was prepared by Technical Committee ISO/TC 69, Subcommittee SC 6, Measurement methods and results. This second edition cancels and replaces the first edition (IS 5725-4:1994), which has been technically revised. ‘The main changes compared to the previous edition are as follows: — clearly recognizing the requirements of the accepted reference values used in bias evaluation experiments and introducing the uncertainties of the accepted reference values, — changing examples with a currently used measurement method. Alist of all parts in the ISO 5725 series can be found on the ISO website. Any feedback or questions on this document should be directed to the user’s national standards body. A complete listing of these bodies can be found at [Link]/[Link]. iv © 150 2020 ~All rights reservedISO 5725-4:2020(E) Introduction ISO 5725 uses two terms, “trueness” and “precision”, to describe the accuracy of a measurement method. “Trueness” refers to the closeness of agreement between the expectation of a measurement result and a true value. “Precision” refers to the closeness of agreement between independent measurement results obtained under stipulated conditions. General consideration of these quantities is given in ISO 5725-1 and so is not repeated in this document. ISO 5725-1 should be read in conjunction with all other parts of ISO 5725, including this document, because it gives the underlying definitions and general principles. ‘The “trueness” of a measurement method is of interest when it is possible to conceive of a true value for the property being measured. Although the true value cannot be known exactly, it can be possible to have an accepted reference value for the property being measured; for example, if suitable reference materials or measurement standards are available, or if the accepted reference value can be established by reference to another measurement method or by preparation of a known sample. The trueness of the measurement method can be investigated by comparing the accepted reference value with the level of the results given by the measurement method. Trueness is normally expressed in terms of bias. Bias can arise, for example, in chemical analysis if the measurement method fails to extract all of an clement, or if the presence of one element interferes with the determination of another. ‘Two measures of trueness are of interest and both are considered in this document. a) Bias of the measurement method: where there is a possibility that the measurement method can give rise to a bias, which persists wherever and whenever the measurement is done, then it is of interest to investigate the “bias of the measurement method’. This requires an experiment involving many laboratories. b) Laboratory bias: measurements within a single laboratory can reveal the “laboratory bias” (as defined in ISO 5725-1). If it is proposed to undertake an experiment to estimate laboratory bias, then it should be realized that the estimate is valid only at the time of the experiment and at the investigated level(§) for the property. Further regular testing is required to show that the laboratory bias does not vary; the method described in ISO 5725-6 can be used for this. @ 180 2020 ~ Al rights reserved vINTERNATIONAL STANDARD ISO S725-4:2020(E) Accuracy (trueness and precision) of measurement methods and results — Part 4: Basic methods for the determination of the trueness of a standard measurement method 1 Scope 1.4 This document — specifies basic methods for estimating the bias of a measurement method and the laboratory bias when a measurement method is applied; — provides a practical approach of abasic method for routine use in estimating the bias of measurement methods and laboratory bias; — provides a brief guidance to all personnel concerned with designing, performing or analysing the results of the measurements for estimating bias. 1.2 Itis concerned exclusively with measurement methods which yield measurements on a continuous scale and give a single value as the measurement result, although the single value can be the outcome of a calculation from a set of observations. 1.3. This document applies when the measurement method has been standardized and all measurements are carried out according to that measurement method. NOTE In ISO/IEC Guide 99:2007(VIM), “measurement procedure” (2.6) isan analogous term related to the term “measurement method” used in this document. 1.4 This document applies only if an accepted reference value can be established to substitute the true value by using the value, for example: — ofa suitable reference material; — ofa suitable measurement standard; — referring to a suitable measurement method; — ofa suitable prepared known sample. 1.5 This document applies only to the cases where it is sufficient to estimate bias on one property at a time, It is not applicable if the bias in the measurement of one property is affected by the level of any other property (Le. it does not consider interferences by any influencing quantity). Comparison of the trueness of two-measurement methods is considered in ISO 5725-6. 2 Normative references The following documents are referred to in the text in such a way that some or all of their content constitutes requirements of this document. For dated references, only the edition cited applies. For undated references, the latest edition of the referenced document (including any amendments) applies. @ 180 2020 ~ Al rights reserved 1ISO 5725-4:2020(E) ISO 3534-1, Statistics — Vocabulary and symbols — Part 1: General statistical terms and terms used in probability ISO 3534-2, Statistics — Vocabulary and symbols — Part 2: Applied statistics ISO 5725-1, Accuracy (trueness and precision) of measurement methods and results — Part 1: General principles and definitions ISO 5725-2, Accuracy (trueness and precision) of measurement methods and results — Part 2: Basic method for the determination of repeatability and reproducibility of a standard measurement method ISO Guide 30, Reference materials — Selected terms and definitions 3 Terms and definitions For the purposes of this document, the terms and definitions given in ISO 3534-1, ISO 3534-2, ISO 5725-1 and ISO Guide 30 apply. 1SO and IEC maintain terminological databases for use in standardization at the following addresses: — 180 Online browsing platform: available at [Link] —_ IEC Electropedia: available at [Link] 4 Symbols A Factor used to calculate the measurement uncertainty of an estimate B Laboratory component of bias G06" ‘Test statistics Gergen Cater Critical values for statistical tests e Random error occurring in every measurement under repeatability conditions G Grubbs' test statistic h Mandel's between-laboratory consistency test statistic k Mangdel's within-laboratory consistency test statistic. a Number of measurement results obtained in one laboratory at one level of the property being measured (i.e. per cell) Number of laboratories participating in the interlaboratory experiment P Probability s Estimate of a standard deviation u Standard measurement uncertainty; quantile of the standard normal distribution y Measurement result y Average of the measurement results y Grand mean of the measurement results « Significance level (cis assumed to be 0,05 in this document) 2 © 150 2020 llrightsreservedISO 5725-4:2020(E) B ‘Type Il error probability Cumulative distribution function of the standard normal distribution Ratio of the reproducibility standard deviation to the repeatability standard y deviation (a9/0,) 6 Bias of the measurement method under investigation 6 Estimate of the bias of the measurement method under investigation 4 Laboratory bias 4 Estimate of the laboratory bias u Accepted reference value of a property being measured v Number of degrees of freedom ° True value ofa standard deviation wv) P-quantile of the y2-distribution with v degrees of freedom Subscripts i Identifier for a participating laboratory; identifier for an individual laboratory (Inner Taboratory) k Identifier for a particular measurement result from a laboratory L Between-laboratory (interlaboratory) m Identifier for detectable bias P Probability r Repeatability R Reproducibility y Identifier for the measurement result 0 Identifier for the accepted reference value 5 Determination of the bias of a standard measurement method by an interlaboratory experiment 5.1 Experimental design considerations 5.1.1 Objective ‘The objective of the experiment is to estimate the value of the bias of the measurement method and to determine when it is statistically significant. When the bias is found to be statistically insignificant, then the objective is to determine the maximum absolute value of bias that can remain, with a certain probability, undetected by the results of the experiment. @ 180 2020 ~ Al rights reserved 3ISO 5725-4:2020(E) 5.1.2 Layout of the experiment ‘The layout of the experiment is almost the same as that for a precision experiment, as described in 180 5725-2. The differences are — the number of participating laborator the requirements given in §.3, and 's and the number of measurement results shall also satisfy — there is an additional requirement, given in 5.4, to use an accepted reference value of the property being measured. 5.1.3 Cross-references to ISO 5725-1 and ISO 5725-2 Requirements on experimental design given in ISO 5725-1 and ISO 5725-2 apply. When reading ISO 5725-1 and ISO 5725-2 in this context, "trueness” shall be inserted in place of “precision” or “repeatability and reproducibility” as appropriate. 5.2 The statistical model ‘The basic model of a measurement result, y, can be expressed as VHH++Bte @ where isthe accepted reference value of a property being measured; 6 isthe bias of the measurement method under investigation; B_ isthe laboratory component of bias; isthe random error occurring in every measurement under repeatability conditions. NOTE __In this document, bias is evaluated at one level at a time; for convenience, the index j, defined in 150 5725-1, for the level of property has been omitted throughout. When all of measurement results are obtained according to the requirements in 5.3 and 5.4 from the sufficient number of participant laboratories and sufficient number of measurements under repeatability conditions in each laboratory by using the same measurement method, the bias of the ‘measurement method, at each level of the property, is estimated by H (2) 5 isanestimate of the bias of the measurement method under investigation; J isthe grand mean of the measurement results from all participant laboratories; KH is the accepted reference value of the property being measured. 5.3 Required number of laboratories and measurements ‘The number of laboratories and the number of measurement results required at each laboratory are interdependent. Guidance on selecting these numbers is given below. Although it is assumed that the laboratory biases can be regarded as draws from an approximately normal distribution, in practice the lance is appropriate for most unimodal di 4 © 150 2020 ~All rights reservedISO 5725-4:2020(E) For the results of an experiment to be able to detect with a high probability a predetermined maximum absolute value of bias, 5, the following formula shall be satisfied: toys @ R184 where A isa factor used in calculating the measurement uncertainty of an estimate of bias (see below); og _ isthe reproducibility standard devi ion of the measurement method; Sn is the predetermined maximum absolute value of bias that the experimenter wishes to de- tect from the results of the experiment; 1,84 isa derived factor (see Annex A). In Formula (3), A is a function of the number of laboratories, the number of measurement results in each laboratory, the reproducibility standard deviation of the measurement method and the measurement uncertainty of the accepted reference value. 4 is given by A=1,96 [ag +z where oa 96, Pw) nly? —1)+1 w oh pn 1,96 is the 0,975-quantile of the standard normal distribution (see Annex A); Ay _ is the ratio of the standard measurement uncertainty of the accepted reference value to the reproducibility standard deviation of the measurement method; is the ratio of the standard deviation of the grand mean in this experiment to the reproducibility standard deviation of the measurement method; u(y). is the standard measurement uncertainty of the accepted reference value; n isthe number of measurement results in each laboratory; p isthe number of participating laboratories; y _ isthe ratio of the reproducibility standard deviation to the repeatability standard deviation. In Formula (4), Ay, A, and y are given respectively by A=u()/oy (5) ON Fon 1=0R/0, ” ©) where o, is the repeatability standard deviation of the measurement method. @ 180 2020 ~ Al rights reserved 5ISO 5725-4:2020(E) If the measurement uncertainty of the accepted reference value is small enough to be neglected, implying 4 $ 0,34, (Le. (1) s 0,34,0,), Formula (4) may be simplified to (8) The values of A calculated by Formula (8) are given in Table 1. Ideally, the choice of the combination of the number of laboratories and the number of replicate ‘measurement results per laboratory should satisfy the requirement described by Formula (3), with the 6,y value predetermined by the experimenter. However, for practical reasons, the choice of the number of laboratories is usually a compromise between the availability of resources and the desire to reduce the value of 6, to a satisfactory extent. If the reproducibility of the measurement method is poor, then it is not practical to achieve a high degree of certainty in the estimate of the bias. When 9, is larger than G, (i.e. y is larger than 1) as is often the case, little is to be gained by obtaining more than n = 2 measurement results per laboratory per level. Table 1 — Values of 4, the factor used in calculating the measurement uncertainty of an estimate of bias in the case when the measurement uncertainty of the accepted reference value is small enough to be neglected [Link] | Value of A calculated by Formula (8) laboratories a mas > 2 a 4 | nez | nea | nea | n=2 | n-3 5 | 062 | 051 | oa | 082 | 060 | 079 | 087 | 086 | 0,86 10 om | 036 | 031 | 058 | 057 | 086 | 061 | 01 | 061 15 036 | 0.29 | 025 | 047 | 046 | 046 | 0,50 | 050 | 050 20 0,31 0,25, 0,22 O41 0,40 040 | 043 0,43 0,43 2 | 028 | 023 | 020 | 037 | 036 | 035 | 039 | 039 | 0.39 30 02s | o21 | oa | 033 | 033 | 032 | 035 | 035 | 036 35 023 | 049 | 047 | 031 | 030 | 030 | 033 | 033 | 033 4 | oz | o1s | oas | 029 | 028 | 028 | 031 | 031 | 031 5.4 Requirements of the accepted reference value 5.4.1 Approaches to assigning the accepted reference value ‘The accepted reference value of the property of interest, 1, shall be reliable and metrologically traceable to an accepted reference by which the true value can be substituted. It refers to the value carried by the material used in this experiment, which was either assigned in another independent study, such as the characterization of the reference material, the calibration of the measurement standard by using suitable calibration procedures and competent laboratories, the assignment by other measurement method (preferably a reference measurement method), or calculated from the property values of the materials used for preparation of the known sample. NOTE Guidance for the characterization and use of reference materials is given in ISO Guide 35 and 150 Guide 33, respectively, Refer to the definition for other measurement standards in ISO/IEC Guide 99 (VIM). 5.4.2 Materials used in the experiment [Link]. The material used in the experiment, whether purchased or prepared, can be a reference ‘material, a measurement standard or a prepared known sample; it shall have the same property as that the standard measurement method is intended to be applied to, e.g. concentration, [Link] The value of the property of interest carried by the material, which has been assigned by any approach listed in 5.4.1, shall be appropriate to the range of values at which the standard measurement 6 © 150 2020 ~All rights reservedISO 5725-4:2020(E) method is intended to be applied. In some cases, it is important to include a series of the materials, each corresponding to a different level of the property, preferably covering the range of values of the measurement method as completely as possible, because the bias of the measurement method can be different at different levels of the property. [Link] The material should have a matrix as close as possible to that to be subjected to the standard measurement method, e.g. carbon in coal or carbon in steel, [Link] The quantity of the materials shall be sufficient for the entire experimental programme, including some in reserve if this is considered necessary, such as for contingencies in transportation and potential re- measurements, 5.4.25 Wherever possible, the materials should have stable property values throughout the experiment. In the case of unstable materials, special instructions on transportation, storage and/or treatment shall be specified, so that the requirements in 5.4.3 can be achieved [Link] Where subdivision of the unit of the material occurs prior to distribution, it shall be performed with care to avoid the introduction of any additional error. Relevant International Standards and other documents on sample division should be consulted. The units shall be selected on a random basis for distribution. If the measurement process is non-destructive, it is possible to give all the laboratories in the interlaboratory experiment the same unit of the material, but this can extend the timeframe of the experiment. [Link] _ In principle, the material should be adequately homogeneous, ie. it should have an acceptable level of heterogeneity. When a material has to be homogenized, this shall be done in the manner most appropriate for that material. When the material to be measured is not sufficiently homogeneous, it is important to prepare the samples in the manner specified in the method, preferably starting with one batch of commercial material for each level of the property value. In particular, the known samples prepared by mixing materials with different known levels of property values in specified proportions, or by spiking a specified proportion of a substance to a matrix material, shall be quantified for their homogeneity, so that the requirements in 5.4.3 can be achieved. 5.4.3. Requirements of measurement uncertainty of the accepted reference value [Link] The standard measurement uncertainty of the accepted reference value, cited or derived from that in relevant documents accompanying the material or estimated for this experiment, shall be controlled to be as low as possible. If it is lower than 0,34,0p, the effect of the accepted reference value on the measurement uncertainty of the estimate of the bias may be neglected. If itis higher than Ayop, two strategies should be considered: — increasing p or both p and n, so that the value of A is reduced to such a level that the predetermined value of bias can be detected with a high probability; — choosing other materials which carry an accepted reference value meeting this requirement. [Link] The standard measurement uncertainty of the accepted reference value consists of, in principle, components arising from characterization (or calibration, or preparation) procedures, heterogeneity and instability, and the measurement uncertainty arising from bias. However, for historical reasons, sometimes the uncertainties arising from heterogeneity, instability and/or bias were not included (maybe separately stated) in the certificate of reference material or measurement standard. They should be estimated by additional studies (or combined with the stated measurement uncertainty), unless there is reliable evidence that they are negligible. [Link] Contribution of the heterogeneity to the measurement uncertainty can depend on the mass or volume used in the measurement. If the heterogeneity measurement uncertainty is the dominant component of the measurement uncertainty of the accepted reference value, the mass or volume used @ 180 2020 ~ Al rights reserved 7ISO 5725-4:2020(E) in the homogeneity check experiment shall be no larger than that to which the measurement method is applied. Otherwise the heterogeneity measurement uncertainty shall he re-assessed based on the latter mass or volume, and the measurement uncertainty of the accepted reference value shall be re-calculated. 5.5 Carrying out the experiment 5.5.1 Evaluation of precision [Link] The precision experiment shall be carried out according to ISO 5725-2. [Link] The measurement results shall be treated as described in ISO 5725-2, where the precision of the measurement method is expressed in terms of s, (estimate of the repeatability standard deviation) and sp (estimate of the reproducibility standard deviation). [Link] The estimate s? of the repeatability variance and the estimate s? of the reproducibility variance are calculated by Formulae (9) to (13), provided that the number (n) of measurement results in each laboratory is equal. If this is not true, the formulae for s, and sy given in ISO 5725-2 shall be used, @ (10) ay 12) a3) where s? and J, are respectively the variance and the average of n measurement results y,, obtained in laboratory i Outliers shall be investigated according to the procedures given in ISO 5725-2. 5.5.2 Check of precision [Link] If the repeatability standard deviation and the reproducibility standard deviation of the standard measurement method have not been previously determined in accordance with ISO 5725-2, s, and sp are considered to be the best estimates of those. 8 © 150 2020 llrightsreservedISO 5725-4:2020(E) If the repeatability standard deviation of the standard measurement method, o,, has been ed in accordance with ISO 5725-2, <2 can be assessed by computing the ratio Cast lop (a4) The test statistic C is compared with the critical value Coit Mina) VY where 72, ,)(¥) is the (1 - @)-quantile of the x? distribution with v= p(a ~ 1) degrees of freedom. a) If C< Cougs S2 isnot significantly larger than a2. b) IFC> C,.¢: 2 is significantly larger than 0? . In the former case, the repeatability standard deviation, ,, is to be used for the assessment of the bias of the measurement method. In the latter case, itis necessary to investigate the causes of the discrepancy and possibly to repeat the experiment prior to proceeding further. NOTE _ Othersstatistical techniques, such as two-side F-test, have been suggested for the check of repeatability and reproducibility standard deviation, it is duty of the panel (refer to Clause 7) to decide which technique is suitable, [Link] Ifthe reproducibility standard deviation, op, and the repeatability standard deviation, 0,, of the standard measurement method have been determined in accordance with ISO 5725-2, sg can be assessed indirectly by computing the ratio (15) where 7. (V) is the (1 ~ @)-quantile ofthe x2 distribution with v = p~ 1 degrees of freedom. a) If C'SCii¢? 5R-(1-1/n) $2 is not significantly larger than 0% -(1-1/n)o?. b) IF C’>Ci: 98-(1-1/n) 2 is significantly larger than of-(1-1/n)o2. In the former case, the repeatability standard deviation, ¢,, and the reproducibility standard deviation, oq, are to be used for the assessment of the trueness of the measurement method. In the latter case, a careful examination of the working conditions of each laboratory shall be carried out before the assessment of the bias of the standard measurement method is undertaken. As a result, the experiment can have to be repeated to yield the expected precision values. NOTE It can appear that some laboratories did not use the required equipment or did not work according to the specified conditions. In chemical analysis, problems can arise from, for example, insufficient control of temperature, moisture, presence of contaminants, etc @ 180 2020 ~ Al rights reserved 9ISO 5725-4:2020(E) 5.5.3. Estimation of the s of the standard measurement method [Link] The estimate of the bias is given by 5=F-n a6) where 6 can be positive or negative. [Link] The variation of the estimate of the bias of the measurement method is due to the variation in the results of the measurement process and the variation of the accepted reference value. In the case of known precision values, the standard deviation of the estimate of the bias is a7) or in the case of unknown precision values, as 2 2 [2 (1-1/1) T(t 1/a) ae (i) (as) P [Link] An approximate 0,95 confidence interval for the bias of the measurement method can be computed as [6-40,, 8+Ac, ] (19) where A is as given in Formula (4). If on is unknown, its estimate sp shall be used instead, and A shall be computed with y= sp/s,rather than Formula (7). Ifthis confidence interval contains the value zero, the bias of the measurement method is insignificant at the significance level «= 0,05; otherwise itis significant. 5.5.4 Example ‘An example for the bias assessment by an inter-laboratory experiment is given in Annex B. 6 Determination of the laboratory bias of one laboratory using a standard measurement method 6.1 Experimental design considerations 6.11 Objective The objective of the experiment is to estimate the value of the bias of a particular laboratory and to determine if it is statistically significant. If the bias is found to be statistically insignificant, then the objective is to determine the value of the maximum bias that can remain, with a certain probability, undetected by the results of the experiment. 6.1.2 Layout of the experiment [Link] An interlaboratory precision experiment, in accordance with ISO 5725-2, has established the repeatability standard deviation of the method before this experiment for estimating the laboratory bias. 10 © 180 2020 - All rights reservedISO 5725-4:2020(E) [Link] The experiment shall conform strictly to the standard method and measurements shall be carried out under repeatability conditions. If the repeatability of the measurement method is poor, then it cannot be practical to achieve small measurement uncertainty in the estimate of the laboratory bias. [Link] The layout of the experiment consists of the measurements required of one laboratory in a precision experiment as described in ISO 5725-2. Apart from the restriction to a single laboratory, the only substantial difference is the additional requirement to use an accepted reference value. The number of measurement results shall satisfy the requirements given in 6.3. 6.1.3 Cross-references to ISO 5725-1 and ISO 5725-2 When reading ISO 5725-1 and ISO 5725-2 in this context, “trueness” should be inserted in place of “precision” or “repeatability and reproducibility” as appropriate, in ISO 5725-2, the number of laboratories is p = 1, and it can be convenient for one person to combine the roles of “executive” and "supervisor". 6.2 The statistical model The laboratory bias, 4, is given by A=6+B (20) So, the model can be written youtate ey 6.3 Number of measurement results ‘The measurement uncertainty of the estimate of the laboratory bias depends on the repeatability of the measurement method, on the number of measurement results obtained and on the measurement uncertainty of the accepted reference value used. For the results of an experiment to be able to detect with a high probability (see Annex A), a predetermined absolute value of bias and the number of measurement results shall satisfy the following formula: Ao, < 22) 1,84 2 where A, isa factor used in calculating the measurement uncertainty of an estimate of laboratory. bias; ;, _ is the repeatability standard deviation of the measurement method, and 4, _ is the predetermined absolute value of laboratory bias that the experimenter wishes to detect from the results of the experiment; 1,84 isa derived factor (see Annex A) @ 180 2020 ~ Al rights reserved WISO 5725-4:2020(E) In Formula (22), A;is a function of the number of measurement results and the measurement uncertainty of the accepted reference value. 4; is given by A=1,96; 2 Alp (23) where 1,96. is the 0,975-quantile of the standard normal distribution (see Annex A); n the number of measurement results; Aj _ is the ratio of standard measurement uncertainty of the accepted reference value to the repeatability standard deviation of the measurement method, Ajp is given by Aig =4(H)/0, where u(y) is the standard measurement uncertainty of the accepted reference value. 64 Requirements of the accepted reference values ‘The requirements given in 5.4 should be achieved except those — related to the transportation conditions; — onthe measurement uncertainty limits of the accepted reference value, which should be controlled as low as possible and can be neglected when u(j1)<0,30, Vn. If the measurement uncertainty of the accepted reference value is higher than o, Vn , e.g. it is hard to obtain a suitable material, the number of measurements should be increased, so that the value of A, can be reduced to the level that the predetermined absolute value of laboratory bias can he detected with a high probability. 6.5 Carrying out the experiment 6.5.1 Check of the within-laboratory standard deviation [Link] Priorto conducting the assessment of laboratory bias, a check of the within-laboratory standard deviation by comparing with the stated repeatability standard deviation of the standard measurement ‘method shall be performed. [Link] Compute the average, J,, of the n measurement results and s,, the estimate of the within- laboratory standard deviation, 6, as follows: De (24) Ci is the average of the n measurement results; nis the number of measurement results; Jie is the kth independent measurement result. 12 © 180 2020 - All rights reservedISO 5725-4:2020(E) (25) where sis the estimate of the within-laboratory standard deviation. [Link] The measurement results shall be scrutinized for outliers using relevant statistical techniques, such as Grubbs’ test, as described in ISO 5725-2. 65.14. If the repeatability standard deviation, o,, of the standard measurement method has been established, the estimate 5; can be assessed by the following procedure: Compute the ratio c*=(5,/6,P (26) and compare the value C" with the critical value Chie = Aina) VV where 72.) (v) is the (1 - a)-quantile of the x? distribution with v= n ~ 1 degrees of freedom, a) C*SCz,, :5)is not significantly larger than 0, b) C7 >Chjp 5/15 significantly larger than 0, In the former case, the repeatability standard deviation of the measurement method, a, is to be used for the assessment of the laboratory bias. In the latter case, consideration should be given to repeating the experiment with verification at all steps that the standard measurement method is properly implemented. 6.5.2 Estimation of the laboratory bias 1 The estimate, A, of the laboratory bias 4 is given by ni @7) 5.2.2 The measurement uncertainty of the estimate of the laboratory bias is due to the variation in the results of the measurement process and the variation of the accepted reference value. In the case of a known repeatability standard deviation, it is expressed by its standard deviation computed as 2 lo: Fae ony ee) (28) ot, in the case of an unknown repeatability standard deviation, as 29) @ 180 2020 ~ Al rights reserved 13ISO 5725-4:2020(E) [Link] The 0,95 confidence interval of the laboratory bias can be computed as [4-4¢,, 4+4¢,] (30) where A/is as given in Formula (23), If¢, is unknown, its estimate, s, shall be used instead. [Link] If this confidence interval contains the value zero, the laboratory bias is insignificant at the significance level a= 0,05; otherwise it is significant. NOTE The laboratory bias is further considered in ISO 5725-6. 7 Report to the panel and decisions to be taken by the panel 7.1 Cross-reference to ISO 5725-2 ‘The personnel and their duties involved i the experiment shall refer to those described in ISO 5725-2. 7.2. Reportby the statistical expert Having completed the statistical analysis, the statistical expert shall write a report to be submitted to the panel. In this report, the following information shall be given: a) a full account of the observations received from the operators and/or supervisors concerning the standard measurement method; b) a full account of the laboratories that have been rejected as outlying laboratories, together with the reasons for their rejection; ) a full account of any stragglers and/or outliers that have been identified, and whether these were explained and corrected, or discarded; 4) a table of the final estimates of appropriate means and precision measures; e) _astatement on whether the bias of the standard measurement method with respect to the accepted reference used is significant; if so, the estimated value of the bias for each level of property shall be reported. 7.3. Decisions by the panel ‘The panel should then discuss the statistical expert's report and take decisions concerning the following questions. a) Are the discordant measurement results, if any, due to defects in the description of the b) 3 a 14 measurement method? What action should be taken with respect to rejected outlying laboratories? Do the results of outlying laboratories and/or the comments received from the operators and supervisors indicate a need to improve the standard measurement method? If so, what are the improvements required? Do the results of the accuracy experiment justify the acceptability of the measurement method for adoption as a standard? What action is to be taken concerning its publication? (e.g, bias can be estimated as negligible (not statistically significant) at only certain levels of the property. These cases can require the panel to suggest reducing the working range of the measurement method. © 150 2020 ~All rights reservedISO 5725-4:2020(E) Annex A (informative) Derivation of formulae A.1 Formulae (3) and (4) The minimum number of laboratories, p, and of measurement results, n, are calculated to satisfy the two following conditions: a) the test should be able to detect that the bias is equal to zero with the probability 1 - a= 09 b) the test should be able to detect a predetermined absolute value of bias, 6,,, with the probability 1-f=095. The first condition is actually developed in [Link], where the confidence interval for the bias of the measurement method, 6, is used to carry out a statistical test of the null hypothesis that the absolute value of bias is equal to zero (Ho:{6| = 0) against the alternative hypothesis that the absolute value of bias is not equal to zero (H;:|6| # 0). An equivalent form of this test is to compare the absolute value of the estimate of the bias of the measurement method with a critical value K, and reject Ho(|6| = 0) if |6|>K [and not reject Ho(|6] = 0) if |8|K J, Kis computed using the requirement that the probability of rejecting Mp, if it is true, shall be equal to the chosen significance level «=0,05: P(\5|>K ||| =0)=a=0,05 (ad) @ 180 2020 ~ Al rights reserved 15ISO 5725-4:2020(E) where ()__ is the cumulative distribution function of the standard normal distribution; up isthe P-quantile of the standard normal distribution; V(8) is the variance of the estimate of the bias of the measurement method. The second condition is that the test should be able to detect the predetermined absolute value of bias, 6, with the probability 1 - f = 0,95: -8,)=0.98 | P(ls|>K| P(l8|sx}l5|=5,, )=, K-6, K=6, -1,645,V (5) (A.2) Equating the two expressions (Formulae A.1 and A.2) for K gives 1,960 ( (1,960+1,645) /V (5) =3,, as) ‘The variance of the estimate of the bias of the measurement method, V (5) , can be derived as follows, v(d)= 7) +1? (uw) (aay 16 © 180 2020 - All rights reservedwhere ul) ISO 5725-4:2020(E) is the standard measurement uncertainty of the accepted reference value obtained from an independent study; V(¥)_ isthe variation of the grand mean obtained from this experiment, of 02 _oR-of 02 _nlog-on/7?)soR/i* po pnp pn pn vpn a k Gf is the between-laboratory variance so that B iL +o? and y= oR/o, Let Ay = uls)/op, A=1,96, (p2eA et, een AS, = A (5) A2 Formulae (22) and (23) ‘These formulae follow immediately if in the preceding derivation (4.1) 5, dy, 6, V(6), A and Ag are replaced by 4, 4, 4, V(4), 4, and Ajq, respectively, and the expression for V(8) is replaced by the expression v(4) @ 180 2020 ~ Al rights reserved o 12 (u) 17ISO 5725-4:2020(E) Annex B (informative) Example of an accuracy experiment B.1 Description of the experiment Anaccuracy experiment on the determination of manganese contentin iron ores by an X-ray fluorescence spectrometry method and using five materials with the accepted reference values (4) is given in Table B.1. (which were not disclosed to the laboratories). Each laboratory received two, randomly selected, bags of sample for each level of manganese content and performed (under repeatability conditions) duplicate analyses on each bag. The purpose of the two-bag system was to confirm absence of the between-bag, variation. The analysis was performed such that in the case where absence of between-bag variation is confirmed, the four analytical results can be considered as replicates under repeatability conditions. Analysis of the results showed that the between-bag variation was indeed insignificant; the materials were regarded to be homogeneous. Thus, results from each laboratory can be considered as replicates under repeatability conditions. The analytical results are listed in Table 8.2. The laboratory means and variances for each of the five materials are listed in Table B.3. B.2 Precision assessment ‘To assess the precision of the analytical method, the data were analysed by the procedure described in ISO 5725-2. The measurement results for concentration level 1, 2, 3, 4 and 5 of manganese content are shown in Figure B.1, Figure B.2, Figure B.3, Figure B4 and Figure B.5. ‘The stragglers and outliers for both Cochran's and Grubbs' tests were identified and are listed in ‘Table B4. The boxed points in Figures B.1 to B.5 signify that the measurement results were identified as stragglers or outliers. Table B.4 shows that two sets of results from two laboratories (Lab. 3 and 7) at two levels of manganese content were identified as outliers by Cochran's test; and one set of results from one laboratory (Lab. 1) at one level of manganese content was identified as stragglers by Grubbs’ test. ‘The Mandel’s h and Mandel’s k values are shown in Figures B.6 and Figure B.. The Mandel’s h values (Figure B.6) show clearly that laboratory 1 gets somewhat lower results at level 2, and they were also identified by Grubbs’ test as stragglers. The Mandel's k values (igure B.7) show that laboratories 3. and 7 tend to get much larger within-laboratory variation than the others respectively at the content level 1 and level 5. Appropriate action should be taken by investigating these laboratories, or, if necessary, by tightening the protocol of the measurement method. For the analysis, it was decided to discard the outliers identified by Cochran's test; ie. the data at content levels 1 of laboratory 3 and the data at content level 5 of laboratory 7. ‘The repeatability and the reproducibility standard deviations were then computed excluding those data that were identified. The results of this computation are summarized in Table B.5 and plotted against the level in Figure B.8. Figure B.8 shows that a linear function seems to be an appropriate relationship between the precisions (repeatability standard deviation, s,, or reproducibility standard deviation, sp) and content levels, m. The linear regression formulae of the repeatability and reproducibility standard deviations versus content levels are: ,009 25m+0,001 15 Sp =0,018 81m+0,002 02 18 © 180 2020 - All rights reservedB.3. Trueness assessment ISO 5725-4:2020(E) The trueness of the measurement method was assessed by computing the 0,95 confidence intervals of the bias of the measurement method using Formula (19) and comparing them with zero (Table B.5). Since at all content levels these confidence intervals contain the value zero, the bias of this measurement method is insignificant at the significance level a= 0,05. B.4. Further analysis Further information can be extracted from the data by carrying out supplementary analyses such as a regression analysis o! versus Table B-1 — Manganese content in iron ores: accepted reference values [Content level 1 2 3 4 5 [Accepted reference value u (4) 280 [0127 | 0403 | 0,650 0,80 [Expanded measurement uncertainty, Uss(%) | 00014 | 00039 | 0.0066 | 0.0092 | 001 [Standard measurement uncertainty, u(u)(%6)__| 0,007 | 0,0020 | 00033 | 0.0046 | 0,0050 ‘Table B.2 — Manganese content in iron ores: analytical results as percentage Mn Unit: % Lab.| Bottle’ Content level No. | No. 2 4 5 1 | 1 [0.0249 00259 [01181 041185 | 0.4127 0.4150 | 0.6898 0.6826 | 0.8214 0.8189 2 [00249 0,0246| 01177 0178 | 04139 04155 | 06839 0,6903 | 0.8283 0,8249 2 | 1 [00316 0.0313 | 0.1352 0.1350 [0.3975 04015 | 0,6603 0,6665 | 0,7820 _0,7876 2 | 00308 0,0315 | 01354 0135404024 04009 | 06494 0,6566 | 0,7887 _0,7867 3 | 1 [0.0222 00224 [01305 01302 [0.4006 0.4004] 06598 06604 | 07910 0.7908 2 [00271 00273] 01303 01301 | 0.4001 04003 | 0.6597 06603 | 0.7905 0.7909 4 | 1 [00271 00290 | 01283 01277 | 04087 0.4072 | 0,6603 06692 | 0,8046 0,8022 2 [00288 0.0276| 01298 0.1282 [04042 04085 | 06632 0.6632 [0.8019 0,8028 S | 1 [6.0271 00271 | 01286 01286 | 0,3957 0,3065 | 0,659.8 06613 | 0,7830 0,781.4 2_| 00271 0,0271| 01293 0.1293 | 0,3957 03957 | 0,6544 0.6552 | 0,782.2 _0,7830 6 | 1 [0.0244 00267 | 01279 01303 | 04054 0.4043 | 0,6603 06603 | 0,7954 0,7872 2 [00251 0.0252 | 01279 0.1284 | 0.4067 0.4030 | 0.6617 0.6608 | 0,7916 0,7941 7 | 1 [0.0269 00283 | 01288 01262 [0.3878 0,3833| 06418 06341 | 08302 0,7994 2 | 00270 0.0260 | 0.1243 0.1284 | 0,3867 0,3801 | 0.6372 0,635 4 | 0.8008 0.8315 8 | 1 [0.0272 00263 | 01271 01295 [0.3900 0.4016 | 0.6420 06416 | 0,8250 0.8319 2 [00279 0,0265| 01242 04286 [03055 03015 | 06352 06325 | 0.8151 0,8202 9 | 1 [00268 0027201298 01301 | 0.4004 0.4054 | 0,6685 06749 | 0,7890 0,7903 2 | 00274 0.0275 | 01297 0.1302 [0.4004 04030 | 0.6617 0.6517 | 0.7859 0,788.4 10 | 1 [00293 0,0304|0,1338 01312 [04044 0.4047 | 0,6591 0,6620 | 0,7903 0.7868 2 | 00292 01337 01308 [04001 04081 | 0,6491 06538 | 0,7903 0,7869 ai [1 [0.0311 0.0306 | 0.1336 0.1355 | 04081 0.4084 0.6770 0,6628 | 0,7962 0,7969 2 [00304 00294 | 01352 0.1359 [0.4074 0.4068 | 06765 06701 | 0,7906 0,8038 12 [1 [00259 00263] 01325 01277 | 0.4100 04127 | 06397 0,6403| 0.7985 0.8037 2 [00250 00257 | 01297 01309 [0.4003 0.4077 | 06413 0.6418 | 0,8156 0,8127 © 150 2020 - A rights reserved 19ISO 5725-4:2020(E) Table B.3 — Manganese content in iron ores: laboratory means and laboratory variances Tab. Content level No. 1 | 2 3 4 5 [Laboratory mean (Unit: %) 1 0,025 1 0.1180 04143 0,6867 0,823 4 2 0,031.3 01353 04006 0,658.2 0,786 3 3 0.0248 01303 04004 0,6601 07908 4 0,028 1 0,285 04072 0,6640 0,029 5 0.0271 041290 03959 0.6577 07824 6 0,025 4 01286 04049 0,6608 07921 7 00271 01269 03850 06371 08155 8 0,0270 01274 03947 0,6378 0,825 3 9 0.0272 0,130.0 04023 0,6642 07884 10 0,029 8 01324 04043 0,6560 0,788 6 i 0,030 4 04351 04077 06716 0.7969 12 0,025 7 01302 04077 06408 0,807.6 Laboratory variance [Uni 1 3.23» 107 129x107 155% 10-6 1.57 «105 168x105) Z 1,27» 107 3,67 x 10-8 4,58 «10-6 5d x 10> 8,70 x 10-6 3 02 « 10° 2,92 x10 4,33 «10 1,23 107 4,67 x 10% 4 849» 107 8,20 «107 4,31 x 10-6 140% 10-5 1,46 «10-6 5 0 1,63 « 107 1,60 «107 1,15 «10-5 5,87 107 6 9,37 «107 4,30 x 10-6 2,48 x 10-6 436107 41,30 * 10-5 7 897x107 4,37 x 10-6 1.61 «105 1,43 « 10-5 3,16 x 10-* 8 5,29 107 5,39 « 10-6 2,69 10 2,23 «10 5,43 x10 2 958x108 5,67 10% 3,77 x 10-6 9,85 10 3,41 x 10-6 10 350% 107 2,85 x 10-6 1,07 * 10 3.27105 3,97 10-6 a 5,09 107 1,02 « 10-6 5,16 x 107 443 «10> 2,98 « 105 12 2,96» 107 4,09 10-6 2,83 « 105 9,02 «107 6,27 x10 ‘Table B.t — Manganese content in iron ores: outliers and stragglers Content level Lab Calculated statistic? Critical value? List of outliers («= 0,01) 1 620 Coil4.t2) = 0,392 5 619 Coos (4.A2) = 0,392 List of stragglers («= 0,05) 2 1 G1=2,531 Gy gg(42) = 2AIZ, Gly y (12) = 2,636 [5 c=Cochran’s test; @1= Grubbs’ test for one smallest outlying observation ‘Table B.5 — Manganese content in iron ores: estimation of repeatability and reproducibility standard deviations and bias of the measurement method Content level 1 2 3 4 5 + 4 4 4 4 lp it 12 12 12 i 20 © 150 2020 ~All rights reservedISO 5725-4:2020(E) Table B.5 (continued) Content level 1 2 3 4 5 5% 00276 01293 0021 0,657 9 07986 5% 0,001 16 0,002 23 0,005 04 0,008 70 0,007 28 se % 0,002 29 0,004.85 0,008 79 0,016 12 0,015 97 ry 1.98 27, 475) 1.85 219) lay 02710 0.2648 0,2506 02552 02770 layin 9% 0,000 62 0,001 28 0,002 20 000412 0,004 42 034 0,081 3 0,079.4 0,075.2 0,076.6 0,083 1 O.3A,s5 % 0,000 19 0,00039 0,000 66 0,001.23 0,001 33 a 0,028 0,127 0.403 0,650 0,80 a0, % 0,0007 0,0020 0,003 3 0.0086 0,005 0 lA 0,305 9 04018 03753 0,285 3 03131 la 0,801 1 09432 0,884 6 0,750.2 0,819.4 las, 9% 0,001 83 0,004 58 0,007 78 0,012 10 0,013 08 $.% 0,000 4 0.0023 0,009 0,007 9 10014 6 ~Asp.% 0,002.2 0023 -0,0087 -0,0042 014 5 6 + As _% 0,0015 0,006 9 0,006 8 00200 0117 m 0,034 0,032 0.03 0.028 0.026 0.024 0,022 002 Key m manganese content in percent L laboratory No. NOTE Boxed points signify that the measurement results were identified as outliers by Cochran's test. Figure B.1 — Manganese content in iron ores — Measurement results at content level 1 (@ 180 2020 Al rights reserved 2ISO 5725-4:2020(E) 0,135 013 o125 Key ‘m manganese content in percent L__ laboratory No. NOTE _ Boxed points signify that the measurement results were identified as stragglers by Grubbs’ test for one outlying observation. Figure B.2 — Manganese content in iron ores — Measurement results at content level 2 mh oa2 | 0415, oat 0.405, o4 0,395, 0,39 0,385, 0,38 0375 Key m manganese content in percent L laboratory No. Figure B.3 — Manganese content in iron ores — Measurement results at content level 3 22 © 180 2020 - All rights reservedISO 5725-4:2020(E) 07 0,69 0,68 0,67 0,66 0,65 0,64 0,63 0,62 Key m manganese content in percent L laboratory No. Figure B.4 — Manganese content in iron ores — Measurement results at content level 4 oss 083 092 ost og 079 078 o77 Key m manganese content, % Laboratory No. NOTE Boxed points signify that the measurement results were identified as outliers by Cochran's test. Figure B.5 — Manganese content in iron ores — Measurement results at content level 5 @ 180 2020 ~ Al rights reserved 23ISO 5725-4:2020(E) tne yah 1h Mandel’s h statistic L_ laboratory No. Figure B.6 — Manganese content in iron ores — Mandel’s h values grouped by laboratories k ; : ll Lh \, if TA at A Mh Key k Mandel’s k statistic L laboratory No. igure B.7 — Manganese content in iron ores — Mandel's k values grouped by laboratories 24 © 180 2020 ~All rights reservedISO 5725-4:2020(E) oo18 0,016 0014 0012 01 0,008 0,006 0,004 0,002 02 03 04 05 06 07 08 09 Key standard deviation, % m manganese content, % sg _ estimated reproducibility standard deviation, 5, estimated repeatability st tandard deviation Figure B.8 — Manganese content in iron ores — Repeatability and reproducibility standard @ 180 2020 ~ Al rights reserved deviations as linear functions of the content m 25ISO 5725-4:2020(E) i) [2] BI (4) [5] [6] 7 26 Bibliography 1S0 3534-3, Statistics — Vocabulary and symbols — Part 3: Design of experiments ISO §725-3, Accuracy (trueness and precision) of measurement methods and results — Part 3: Intermediate measures of the precision ofa standard measurement method ISO 5725-5, Accuracy (trueness and precision) of measurement methods and results — Part 5: Alternative methods for the determination of the precision of a standard measurement method ISO 5725-6, Accuracy (trueness and precision) of measurement methods and results — Part 6: Use in practice of accuracy values ISO Guide 33, Reference materials — Good practice in using reference materials ISO Guide 35, Reference materials — Guidance for characterization and assessment of homogeneity and stability ISO/IEC Guide 99:2007, International vocabulary of metrology — Basic and general concepts and associated terms (VIM) © 150 2020 ~All rights reservedISO 5725-4:2020(E) ICS 03.120. Price based on 26 pages 7.020 (© 180 2020 - All rights reserved