Sie sind auf Seite 1von 35

Disclosure to Promote the Right To Information

Whereas the Parliament of India has set out to provide a practical regime of right to
information for citizens to secure access to information under the control of public authorities,
in order to promote transparency and accountability in the working of every public authority,
and whereas the attached publication of the Bureau of Indian Standards is of particular interest
to the public, particularly disadvantaged communities and those engaged in the pursuit of
education and knowledge, the attached public safety standard is made available to promote the
timely dissemination of this information in an accurate manner to the public.

1 +, 1 + 01 ' 5
Mazdoor Kisan Shakti Sangathan Jawaharlal Nehru
The Right to Information, The Right to Live Step Out From the Old to the New

IS 15393-3 (2003): Accuracy (Trueness and Precision) of


Measurement Methods and Results, Part 3: Intermediate
Measures of the Precision of a Standard Measurement Method
[PGD 25: Engineering Metrology]

! $ ' +-
Satyanarayan Gangaram Pitroda
Invent a New India Using Knowledge

! > 0 B


BharthariNtiatakam
Knowledge is such a treasure which cannot be stolen
IS 15393 (Part 3): 2003
ISO 5725-3:1994

W-F?a%7
m
m~amnmw-iiq?l
Wmkl-r( W-lia @m-f@G)

Indian Standard
ACCURACY ( TRUENESS AND PRECISION ) OF
MEASUREMENT METHODS AND RESULTS
PART 3 INTERMEDIATE MEASURES OF THE PRECIStON
OF A STANDARD MEASUREMENT METHOD

ICS 17.020; 03.120.30

@ BIS 2003

BUREAU OF INDIAN STANDARDS


MANAK BHAVAN, 9 BAHADUR SHAH ZAFAR MARG
NEW DELHI 110002

September 2003
Price Group 10
Engineering Metrology Sectional Committee, BP 25

NATIONAL FOREWORD

This Indian Standard ( Part 3 ) which is identical with ISO 5725-3:1994 Accuracy ( trueness and
precision ) of measurement methods and rfkults Part 3: Intermediate measures of the precision of
a standard measurement method issued by the International Organization for Standardization ( ISO )
was adopted by the Bureau 01 Indian Standard on the recommendations of the Engineering Metrology
Sectional Committee and approval of the Basic and Production Engineering Division Council,

This standard specifies the four intermediate precision measures due to changes in observation conditions
( time, calibration, equipments and operator) within a laboratory.

Many different factors may contribute to the variability of results from a measurement method including
operator, equipment used, calibration of equipment, environment ( temperature, humidity air-pollution,
etc ); batch of a reagent and time elapsed between measurements.

The text of the ISO Standard has been approved as suitable for publication as an Indian Standard
without deviations. In this adopted standard -certain conventions are, however, not identical to those
used in Indian Standards. Attention is particularly drawn to the following:

a) Wherever the words International Standard appear referring to this standard, they should be
read as Indian Standard.

b) Comma ( , ) has been used as a decimal marker in the International Standards, while in Indian
Standards, the current practice is to use a point ( . ) as the decimal marker.

In the adopted standard, reference appears to the following International Standards for which Indian
Standards also exists. The corresponding Indian Standards which are to be substituted in their place
are listed below along with their degree of equivalence for the editions indicated:

International Standard Corresponding Indian Standard Degree of Equivalence

ISO 5725-1 : 1994 Accuracy IS 15393 ( Part 1 ) :2003 Accuracy Identical


( trueness and precision ) of ( trueness and precision ) of
measurement methods and measurement methods and
results Part 1 : General principles results : Part 1 General principles
and definitions and definitions

ISO 5725-2 : 1994 Accuracy IS 15393 ( Part 2 ) :2003 Accuracy do


( trueness and precision ) of ( trueness and precision ) of
measurement methods and measurement methods and
results Part 2: Basic method for results : Part 2 Basic method for the
the determination of repeatability determination of repeatability and
and reproducibility of a standard reproducibility of a standard
measurement method measurement melhod

This standard ( Part 3 ) covers the intermediate measures of the precision of a standard measurement
method. The other five parts of the standard are listed below:

IS No. Title

IS 15393 ( Part 1 ) : 2003/ Accuracy ( trueness and precision ) of measurement methods and
ISO 5725-1:1994 results : Part 1 General principles and definitions

IS 15393 ( Part 2 ) : 2003/ Accuracy ( trueness and precision ) of measurement methods and
ISO 5725-2:1994 results : Part 2 Basic method for the determination of repeatability and
reproducibility of a standard measurement method

( Continued on third cover)


IS 15393 (Part 3):2003
ISO 5725-3:1994

Introduction

0.1 ISO 5725 uses two terms trueness and precision to describe
the accuracy of a measurement method. Trueness refers to the close-
ness of agreement between the average value of a large number of test
results and the true or accepted reference value. Precision refers to the
closeness of agreement between test results.

0.2 General consideration of these quantities is given in ISO 5725-1 and


so is not repeated here. It is sttessed that ISO 5725-1 should be read in
conjunction with all other parts of ISO 5725 because the underlying defi-
nitions and general principles are given there.

0.3 Many different factors (apati from variations between supposedly


identical specimens) may contribute to the variab~lity of results from a
measurement method, including:

a) the operator;

b) the equipment used;

c) fhe calibration of the equipment;

d) the environment (temperature, humidity, air pollution, etc.);

e) the batch of a reagent;

f) the time elapsed between measurements.

The variability between measurements performed by different operafors


and/or with different equipment will usually be greater than fhe variability
between measurements carried out within a short interval of time by a
single operator using fhe same equipmenf.

0.4 Two conditions of precision, fermed repeatability and reproducibility


conditions, have been found necessary and, for many -pracfical eases,
useful for describing the variability of a measurement method. Under re-
peatability conditions, factors a) to f) in 0.3 are considered constanfs and
do not contribute to the variability, while under reproducibility conditions
they vary and do contribute to the variability of the test results. Thus re-
peafabilify and reproducibility conditions are the fwo extremes of pre-
cision, the first describing the minimum and the second fhe maximum
variability in results. Intermediate conditions between these two extreme
conditions of precision are also conceivable, when one or more of factors

.,
I
IS 15393 (Part 3) :2003
ISO 5725-3:1994

a) to f) are allowed to vary, and are used in certain specified circum-


stances.

Precision is normally expressed in terms of standard deviations.

0.5 This part of ISO 5725 focuses on intermediate precision measures


of a measurement method. Such measures are called intermediate as
their magnitude lies between the two extreme measures of the precision
of a measurement method: repeatability and reproducibility standard de-
viations.

To illustrate the need for such intermediate precision measures, consider


the operation of a present-day laboratory connected with a production
plant involving, for example, a three-shift working system where
measurements are made by different operators on different equipment.
Operators and equipment are then some of the factors that contribute to
the variability in the test results. These factors need to be taken into ac-
count when assessing the precision of the measurement method.

0.6 The intermediate precision measures defined in this part of


ISO 5725 are primarily useful when their estimation is part of a procedure
that aims at developing, .standardizing, or controlling a measurement
method within a Iaboratoty. These measures can also be estimated in a
specially designed interlaboratory study, but their interpretation and appli-
cation then requires caution for reasons explained in 1.3 and 9.1.

0.7 The four factors most likely to influence the precision of a


measurement method are the following.

a) Time: whether the time interval between successive measurements


is short or long.

b) Calibration: whether the same equipment is or is not recalibrated


between successive groups of measurements.

c) Operato~ whether the same or different operators carry out the suc-
cessive measurements.

d) Equipment: whether the same or different equipment (or the same


or different batches of reagents) is used in the measurements.

0.8 It is, therefore, advantageous to introduce the following J4-factor-


different intermediate precision conditions (44 = 1, 2, 3 or 4) to take ac-
count of changes in measurement conditions (time, calibration, operator
and equipment) within a laboratory.

a) 14=1: only one of the four factors is different;

b) M=2: two of the four factors are different;

C) M=3: three of the four factors are different;

d) M=4: all four factors are different.

Different intermediate precision conditions lead to different intermediate


precision standard deviations denoted by S,(~, where the specific con-
ditions are listed within the parentheses. For example, Sluo) is the inter-

ii
IS 15393 (Part 3) :2003
ISO !5725-3 :1994

mediate precision standard deviation with different times (T) and


operators (0).

0.9 For measurements under intermediate precision conditions, one or


more of the factors listed in 0.7 is or are different. Under repeatability
conditions, those factors are assumed to be constant.

The standard deviation of test results obtained under repeatability con-


ditions is generally less than that obtained under the conditions for inter-
mediate precision conditions. Generally in chemical analysis, the standard
deviation under intermediate precision conditions may be two or three
times as large as that under repeatability conditions. It should not, of
course, exceed the reproducibility standard deviation.

As an example, in the determination of copper in copper ore, a


collaborative experiment among 35 laboratories revealed that the standard
deviation under one-factordifferent intermediate precision conditions (op-
erator and equipment the same but time different) was 1,5 times larger
than that under repeatability conditions, both for the electrolytic gravimetry
and NazS20~ titration methods.

...
M
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Indian Standard
ACCURACY ( TRUENESS AND PRECISION ) OF
MEASUREMENT METHODS AND RESULTS
PART 3 INTERMEDIATE MEASURES OF THE PRECISION
OF A STANDARD MEASUREMENT METHOD

1 Scope the outcome of a calculation from a set of obser-


vations.

1.1 This part of ISO 5725 specifies four intermedi- 1.3 The essence of the determination of these in-
ate precision measures due to changes in observation termediate precision measures is that they measure
conditions (time, calibration, operator and equipment) the ability of the measurement method to repeat test
within a laboratory. These intermediate measures can results under the defined conditions.
be established by an experiment within a specific
laboratory or by an interlaboratory experiment. 1.4 The statistical methods developed in this part
of ISO 5725 rely on the premise that one can pool
Furthermore, this part of ISO 5725 information from similar measurement conditions
to obtain more accurate information on the inter-
a) discusses the implications of the definitions of in-
mediate precision measures. This premise is a
termediate precision measures;
powerful one as long as what is claimed as similar
is indeed similar. But it is very difficult for this
b) presents guidance on the interpretation and appli-
premise to hold when intermediate precision meas-
cation of the estimates of intermediate precision
ures are estimated from an interlaboratory study. For
measures in practical situations;
example, controlling the effect of time or of oper-
ator across laboratories in such a way that they are
c) does not provide any measure of the errors in
estimating intermediate precision measures; similar, so that pooling information from different
laboratories makes sense, is very difficult. Thus, using
d) does not concern itself with determining the results from interlaboratory studies on intermediate
trueness of the measurement method itse~, but precision measures requires caution. Within-
does discuss the connections between trueness Iaborator-y studies also rely on this premise, but in
and measurement conditions. such studies it is more likely to be realistic, because
the control and knowledge of the actual effect of a
factor is then more within reach of the analyst.
1.2 This part of ISO 5725 is concerned exclusively
with measurement methods which yield measure- 1.5 There exist other techniques besides the ones
ments on a continuous scale and give a single value described in this part of ISO 5725 to estimate and to
as the test result, although the single value may be verify intermediate precision measures within a lab-

1
IS 15393 (Part 3) :2003
ISO 5725-3:1994

oratory, for example, control charts (see ISO 5725-6). 3 Definitions


This part of ISO 5725 does not claim to describe the
only approach to the estimation of intermediate pre- For the purposes of this part of ISO 5725, the defi-
cision measures within a specific laboratory. nitions given in ISO 3534-1 and ISO 5725-1 apply.

NOTE 1 This part of ISO 5725 refers to designs of ex- The symbols used in ISO 5725 are given in annex A.
periments such as nested designs. Some basic information
is given in annexes B and C. Other references in this area
are given in annex E.

2 Normative references 4 General requirement

The following standards contain provisions which, In order that the measurements are made in the same
through reference in this text, constitute provisions way, the measurement method shall have been
of this part of ISO 5725. At the time of publication, the standardized. All measurements forming part of an
editions indicated were valid. All standards are subject experiment within a specific laborato~ or of an inter-
to revision, and parties to agreements based on this Iaboratoty experiment shall be carried out according
part of ISO 5725 are encouraged to investigate the to that standard.
possibility of applying the most recent editions of the
standards indicated below. Members of IEC and ISO
maintain registers of currently valid International
Standards.
5 Important factors
ISO 3534-1:1993, Statistics Vocabulary and sym-
bols Part 1: Probability and general statistical
terms.
5.1 Four factors (time, calibration, operator and
ISO 5725-1:1994, Accuracy (trueness and precision) equipment) in the measurement conditions within a
of measurement methods and results Part 1: laboratory are considered to make the main contri-
General principles and definitions. butions to the variability of measurements (see
table 1).
ISO 5725-2:1994, Accuracy (trueness and precision)
of measurement methods and results Part 2. Basic
method for the determination of repeatability and
5.2 Measurements made at the same time in-
reproducibility of a standard measurement method.
clude those conducted in as short a time as feasible
ISO Guide 33:1989, Uses of certified reference ma- in order to minimize changes in conditions, such as
terials. environmental conditions, which cannot always be
guaranteed constant. Measurements made at differ-
ISO Guide 35:1989, Certification of reference ma- ent times, that is those carried out at long intervals
terials General and statistical principles. of time, may include effects due to changes in en-
vironmental conditions.

Table 1 Four imDortant factors and their states

Measurementconditionswithin a laboratory
Factor
State 1 (same) I State2 (dtierent)

Time Measurements made at the same Measurements made at different


time times
Calibration No calibration between measure Calibration carried out between
ments measurements
Operator Same operator Different operators
Equipment Same equipment without recali- Different equipment
i bration
I

2
IS 15393 {Part 3) :2003
ISO 5725-3:1994

5.3 Calibration does not refer here to any cali- operatordiffere nt intermediate precision standard
bration required as an integral part of obtaining a test ~eviation, s,(o);
result by the measurement method. It refers to the
calibration process that takes place at regular intervals I[time + operator]-different intermediate precision
between groups of measurements within a labora- standard deviation, quo);
tory.
I[time + operator + equipment]-different inter-
mediate precision standard deviation, slfloE);

5.4 In some operations, the operator may be, in


md many others in a similar fashion.
fact, a team of operators, each of whom performs
some specific part of the procedu~e. In such a case,
the team should be regarded as the operator, and any
change in membership or in the allotment of duties
within the team should be regarded as providing a
6 Statistical model
different operator.

6.1 Basicmodel
5.5 Equipment is often, in fact, sets of equip-
For estimating the accuracy (trueness and precision)
ment, and any change in any significant component
of a measurement method, it is useful to assume that
should be regarded as providing different equipment.
every test result, y, is the sum of three components:
As to what constitutes a significant component,
common sense must prevail. A change of y=m+B+e . . . (1)
thermometer would be considered a significant com-
ponent, but using a slightly different vessel to contain where, for the particular material tested,
a water bath would be considered trivial. A change of
a batch of a reagent should be considered a significant m is the general mean (expectation);
component. It can lead to different equipment or to
B is the laboratory component of bias under re-
a recalibration if such a change is followed by cali-
peatability conditions;
bration.
e is the random error occurring in every
measurement under repeatability conditions.
5.6 Under repeatability conditions, all four factors
A discussion of each of these components, and of
are at state 1 of table 1. For intermediate precision
extensions of this basic model, follows.
conditions, one or more factors are at state 2 of
table 1, and are specified as precision conditions
with M factor(s) different, where Al is the number
of factors at state 2. Under reproducibility conditions, 6.2 General mean, m
results are obtained by different laboratories, so that
not only are all four factors at state 2 but also there
6.2.1 The general mean, m, is the overall mean of
are additional effects due to the differences between
the test results. The value of m obtained in a
laboratories in management and maintenance of the
collaborative study (see ISO 5725-2) depends solely
laboratories, general training levels of operators, and
on the true value and the measurement method,
in stability and checking of test results, etc.
and does not depend on the laboratory, equipment,
operator or time by or at which any test result has
been obtained. The general mean of the particular
5.7 Under intermediate precision conditions with M material measured is called the level of the test; for
factor(s) different, it is necessary to specify which example, specimens of different purities of a chemical
factors are at state 2 of table 1 by means of suffixes, or different materials (e.g. different types of steel) will
for example: correspondto different levels.

time-different intermediate precision standard de- In many situations, the concept of a true value p holds
viation, Sip); good, such as the true concentration of a solution
which is being titrated. The level m is not usually equal
calibration-different intermediate precision stan- to the true value p; the difference (m p) is called the
dard deviation, s,(c); bias of the measurement method.

3
IS 15393 (Part 3):2003
ISO 5725-3:1994

In some situations, the level of the test is exclusively In practice, the objectives of a study and consider-
defined by the measurement method, and the con- ations of the sensitivity of the measurement method
cept of an independent true value does not apply; for will govern the extent to which this model is used. In
example, the Vickers hardness of steel and the many cases, abbreviated forms will suffice.
Micum indices of coke belong to this category. How-
ever, in general, the bias is denoted by 6 (6 = O where
no true value exists), then the general mean m is

6.4 Terms Bo, B(l), B(2),etc.

NOTE 2 Discussion of the bias term d and a description


of trueness experiments are given in ISO 5725-4.
6.4.1 Under repeatability conditions, these terms all
remain constant and add to the bias of the test re-
sults. Under intermediate precision conditions, B. is
6.2.2 When examining the difference between test the fixed effect of the factor(s) that remained the
results obtained by the same measurement method, same (state 1 of table 1), while B(l), B(2), etc. are the
the bias of the measurement method may have no random effects of the factor(s) which vary (state 2 of
influence and can be ignored, unless it is a function table 1). These no longer contribute to the bias, but
of the level of the test. When comparing test results increase the intermediate precision standard deviation
with a value specified in a contract, or a standard so that it becomes larger than the repeatability stan-
value where the contract or specification refers to the dard deviation.
true value p and not to the level of the test m, or when
comparing test results obtained using different
measurement methods, the bias of the measurement
6.4.2 The effects due to differences between oper-
method must be taken into account.
ators include personal habits in operating measure-
ment methods (e.g. in reading graduations on scales,
etc.). Some of these differences should be removable
by standardization of the measurement method, par-
6.3 Term B
ticularly in having a clear and accurate description of
the techniques involved. Even though there is a bias
6.3.1 B is a term representing the deviation of a in the test results obtained by an individual operator,
laboratory, for one or more reasons, from m, irre- that bias is not always constant (e.g. the magnitude
spective of the random error e occurring in every test of the bias will change according to hislher mental
result. Under repeatability conditions in one labora- and/or physical conditions on that day) and the bias
tory, B is considered constant and is called the lab- cannot be corrected or calibrated exactly. The magni-
oratory component of bias. tude of such a bias should be reduced by use of a
clear operation manual and training. Under such cir-
cumstances, the effect of changing operators can be
6.3.2 However, when using a measurement method considered to be of a random nature.
routinely, it is apparent that embodied within an
overall value for B are a large number of effects which
are due, for example, to changes in the operator, the
6.4.3 The effects due to differences between
equipment used, th-e calibration of the equipment, and
equipment include the effects due to different places
the environment (temperature, humidity, air pollution,
of installation, particularly in fluctuations of the indi-
etc.). The statistical model [equation (1)] can then be
cator, etc. Some of the effects due to differences
rewritten in the form:
between equipment can be corrected by exact cali-
bration. Differences due to systematic causes be-
y=m+Bo+B U) +B(2) + ... + e . . . (3)
tween equipment should be corrected by calibration,
and such a procedure should be included in the stan-
or
dard method. For example, a change in the batch of
y=p+f5+ Bo+B(l)+B(2) +...+ e . ..(4) a reagent could be treated that way. An accepted
reference value is needed for this, for which
ISO Guide 33 and ISO Guide 35 shall be consulted.
where B is composed of contributions from variates The remaining effect due to equipment which has
Bo, B(l), B(2) ..:. ,and can account for a number of inter- been calibrated using a reference material is con-
mediate precmon factors. sidered a random effect.

4
IS 15393 (Pati 3):2003
ISO 5725-3:1994

6.4.4 The effects due to time may be caused by 6.5 Errorterm, e


environmental differences, such as changes in room
temperature, humidity, etc. Standardization of en- 6.5.1 This term represents a random error occurring
vironmental conditicsns should be attempted to mini- in every test result and the procedures given
mize these effects. throughout this part of ISO 5725 were developed as-
suming that the distribution of this error variable is
approximately normal, but in practice they work for
most distributions provided that they are unimodal.

6.4.5 The effect of skill or fatigue of an operator may


be considered to be the interaction of operator and 6.5.2 Wkhin a single laboratory, its variance is called
time. The performance of a set of equipment may be the within-laboratory variance and is expressed as
different at the time of the start of its use and after
Var(e) = & . . . (8)
using it for many hours: this is an example of inter-
action of equipment and time. .When the population
of operators is small in number and the population of
sets of equipment even smaller, effects caused by 6.5.3 It may be expected that u~ will have different
these factors may be evaluated as fixed (not random) values in different laboratories due to differences such
effects as in the skills of the operators, but in this part of
ISO 5725 it is assumed that, for a properly standard-
ized measurement method, such differences be-
tween laboratories should be small and that it is
justifiable to establish a common value of within-
6.4.6 The procedures given in ISO 5725-2 are de- Iaboratory variance for all the laboratories using the
veloped assuming that the distribution of laboratory measurement method. This common value, which is
components of bias is approximately normal, but in estimated by the mean of the within-laboratory vari-
practice they work for most distributions provided that ances, is called the repeatability variance and is
these distributions are unimodal. The variance of B is designated by
called the between-laboratory variance, expressed
a:= Var(e) . . . (9)
as

Var(B) = at . . . (5) This mean value is taken over all the laboratories tak-
ing part in the accuracy experiment which remain af-
ter outliers have been excluded.
However, it will also include effects of changes of
operator, equipment, time and environment. From a
precision experiment using different operators,
measurement times, environments, etc., in a nested 7 Choice of measurement conditions
design, intermediate precision variances can be cal-
culated. Var(13) is considered to be composed of in-
7.1 In applying a measurement method, many
dependent contributions of laborato~, operator, day
measurement conditions are conceivable within a
of experiment, environment, etc.
Iaboratov, as follows:

Var(B) = Var(BO) + Var(B(l)) + Var(B(2)) + ...


a) repeatability conditions (four factors constant);

. . . (6)
b) several intermediate precision conditions with one
factor different;
The variances are denoted by

c) several intermediate precision conditions with two


Var(BO) = ti~o)
factors different;
Var(B(l)) = u~l)
d) several intermediate precision conditions with
three factors different;
va@(2)) = 42p etc. . . . (7)

e) intermediate precision conditions with four factors


Var(B) is estimated in practical terms as s: and similar different.
intermediate precision estimates may be obtained
from suitably designed experiments. In the standard for a measurement method, it is not
necessary (or even feasible) to state ail possible pre-

5
IS 15393 (Part 3):2003
ISO 5725-3:1994

cisioi-r measures, although the repeatability standard of factor(s) between each measurement. It is rec-
deviation should always be specified. As regards in- ommended that n should bq at Ieast 15. This may not
termediate precision measures, common commercial be satisfactory for the laboratory, and this method of
practice should indicate the conditions normally en- estimating intermediate precision measures within a
countered, and it should be sufficient to specify only labor&ory cannot be regarded as efficient when
the one suitable intermediate precision measure, to- compared with other procedures. The analysis is
gether with the detailed stipulation of the specific simple, however, and it can be useful for studying
measurement conditions associated with it. The time-different intermediate precision by making suc-
measurement condition factor(s) to be changed cessive measurements on the same sample on suc-
should be carefully defined; in particular, for time- cessive days, or for studymg the effects of calibration
different intermediate precision, a practical mean time between measurements.
interval between successive measurements should
be specified. A graph of (yk Y) versus the measurement number
k, where yk is the kth test result of n replicate test re-
sults and Y is the mean of the n replicate test results,
is recommended to identify potential outliers. A more
7.2 It is assumed that a standardized measurement
formal test of outliers consists of the application of
method will be biased as little as possible, and that
Grubbs test as given in subclause 7.3.4 of
the bias inherent in the method itself should have
ISO 5725-2:1994.
been dealt with by technical means. This part of
ISO 5725, therefore, deals only with the bias coming The estimate of the intermediate precision standard
from the measurement conditions. deviation with M factor(s) different is given by

7.3 A change in the factors of the measurement


conditions Kline, calibration, operator and equipment)
Slo =
/z In
n1
k=l
(yk - ~) . . . (lo)

from repeatability conditions (i.e. from state 1 to 2 of


where symbols denoting the intermediate precision
table 1) will increase the variability of test results.
conditions should appear inside the parentheses.
However, the expectation of the mean of a number
of test results will be less biased than under repeat-
ability conditions. The increase in the standard devi-
ation for the intermediate precision conditions may be 8.2 An alternative method
overcome by not relying on a single test result but by
using the mean of several test results as the final
8.2.1 An alternative method considers r groups of
quoted result.
measurements, each comprising n replicate test re-
sults. For example, within one laboratory, a set of r
materials could each be measured, then the inter-
7.4 Practical considerations in most laboratories, mediate precision factor(s) could be altered and the r
such as the desired precision (standard deviation) of materials remeasured, the procedure being repeated
the final quoted result and the cost of performing the until there are n test results on each of the r materials.
measurements, will govern the number of factors and Each group of n test results shall be obtained on one
the choice of the factor(s) whose changes can be identical sample (or set of presumed identical samples
studied in the standardization of the measurement in the case of destructive testing), but it is not es-
method. sential that the materials be identical, It is only re-
quired that the t materials all belong to the interval of
test levels within which one value of the intermediate,
precision standard deviation with M factor(s) different
8 Within-laboratory study and analysis can be considered to apply. It is recommended that
of intermediate precision measures the value of t(n 1) should be at least 15.

8.1 Simplest approach EXAMPLE

The simplest method of estimating an intermediate One operator performs a single measurement on
precision standard deviation within one laboratory each of the t materials, then this is repeated by a
consists of taking one sample (or, for destructive second operator, and possibly by a third operator, and
testing, one set of presumably identical samples) and so on, allowing an estimate of s,(o) to be calculated.
performing a series of n measurements with a change

6
IS 15393 (Part 3):2003
ISO 5725-3:1994

8.2.2 A graph of (~ jj) versus the material number can be guaranteed only under conditions of a specific
j, where yjk is the k test result on the jth material and operator, equipment or time may not be good enough
~ is the average of the n results on the jth material, is for commercial considerations.
recommended to identify potential outliers. A more
formal test of outliers consists of the application of
Grubbs test as given in subclause 7.3.4 of
ISO 5725-2:1994 either for each group separately or 9 Interlabcwatory study and analysis of
for all m test results combined. intermediate precision measures
The estimate of the intermediate precision standard
deviation with M factor(s) different, S,(), is then given 9.1 Underlying assumptions
by
Estimation of intermediate measures of precision
I
from interlaboratory studies relies on the assumption
that the effect of a particular factor is the same across
all laboratories, so that, for example, changing oper-
ators in one laboratory has the same effect as chang-
ing operators in another laboratory, or that variation
For n = 2 (i.e. two test results on each material), the due to time is the same across all laboratories. If this
formula simplifies to assumption is violated, then the concept of inter-
mediate measures of precision does not make sense,
nor do the techniques proposed in the subsequent
(12) sections to estimate these intermediate measures of
precision. Careful attention to outliers (not necessarily
deletion of outliers) must be paid as this will help in
detecting departure from the assumptions necessary
to pool information from all laboratories. One powerful
8.3 Effect of the measurement conditions on technique to detect potential outliers is to depict the
the final quoted result measurements graphically as a function of the various
levels of the factors or the various laboratories in-
8.3.1 The expectation of Y is different between one eluded in the study.
combination and another of time, calibration, operator
and equipment, even when only one of the four fac-
tors changes. This is a limitation on the usefulness of
9.2 Simplest approach
mean values. In chemical analysis or physical testing,
Y is reported as the final quoted result. In trading raw
If material at q levels is sent to p laboratories who
materials, this final quoted result is often used for
each perform measurements on each of the q levels
quality evaluation of the raw materials and affects the
with a change of intermediate precision factor(s) be-
price of the product to a considerable extent.
tween each of the n measurements, then the analysis
is by the same method of calculation as explained in
EXAMPLE ISO 5725-2, except that an intermediate precision
standard deviation is estimated instead of the repeat-
In the international trading of coal, the size of the
ability standard deviation.
consignment can often exceed 70000 t, and the ash
content is determined finally on a test portion of only
1 g. In a contract stipulating that each difference of
1 ?4. in ash content corresponds to USD 1,5 per tonne 9.3 Nested experiments
of coal, a difference of 1 mg in the weighing of ash
by a chemical balance corresponds to 0,1 YO in ash A further way of estimating intermediate precision
content, or -USD 0,15 per tonne, which for such a measufes is to conduct more sophisticated exper-
consignment amounts to a difference in proceeds of iments. These can be fully-nested or staggered-
USD 10500 (from 0,1 x 1,5 x 70 000). nested experiments (for definitions of these terms,
see ISO 3534-3). The advantage of employing a
nested experimental design is that it is possible, at
8.3.2 Consequently, the final quoted result of one time and in one interlaborato~ experiment, to
chemical analysis or physical testing should be suf- estimate not only repeatability and reproducibility
ficiently precise, highly reliable and, especially, uni- standard deviations but also one or more intermediate
versal and reproducible. A final quoted result which precision standard deviations. There are, however,

7
IS 15393 (Part 3) :2003
ISO 5725-3:1994

certain caveats which must be considered, as will be


explained in 9.8. The subscripts i, j and k suffixed to the data y in
figure 1 a) for the three-factor fully-nested experiment
represent, for example, a laboratory, a day of exper-
9.4 Fully-nested experiment iment and a replication under repeatability conditions,
respectively.
A schematic layout of the fully-nested experiment at
a particular level of the test is given in figure 1. The subscripts i, j, k and f suffixed to the data y in
figure 1 b) for the four-factor fully-nested experiment
By carrying out the three-factor fully-nested exper- represent, for example, a laborato~, a day of exper-
iment collaboratively in several laboratories, one in- iment, an operator and a replication under repeatability
termediate precision measure can be obtained at the conditions, respectively.
same time as the repeatability and reproducibility
standard deviations, i.e. U(O, a(l) and a, can be esti- Analysis of the results of -an n-factor fully-nested ex-
mated. Likewise the four- 1actor fully-nested exper- periment is carried out by the statistical technique
iment can be used to obtain two intermediate analysis of variance (ANOVA) separately for each
precision measures, i.e. U(0), u(l), a(z) and a, can be level of the test, and is described in detail in
estimated. annex B.

FACTOR

O (laboratory) i ---

1 j---

2 (residual) k ---

Y/j/t Yin Yi12 Yi21 Yi22

a) Three-factor ful~y-nested experiment

FACTOR

O (laboratory) i ---

1 J----
1
2 k ---
I
3 (residual} ( ---
I 1
Y ijkl Y /111 Yil12 Yi121 Yi122 Yi211 Yi212 Yi221 Yi222

b) Four-factor futly-nested experiment

Figure 1 Schematic layouts for three-factor and four-factor fully-nested experiments

8
tS 15393 (Part 3):2003
1s0 5725-3:1994

9.5 Staggered-nested experiment should be in the lowest ranks, the lowest factor being
considered as a residual variation. For example, in a
A schematic layout of the staggered-nested exper- four-factor design such as illustrated in figure 1 b and
iment at a particular level of the test is given in figure 2, factor O could be the Iaboratoty, factor 1 the
figure 2. operator, factor 2 the day on which the measurement
is carried out, and factor 3 the replication. This may
not seem important in the case of the fully-nested
FACTOR experiment due to its symmetty.

O (Laboratory) i ---
I 9.7 Comparison of the nested design with
1
the procedure given in ISO 5725-2
1 1
2

3 (residual) j ---
I 1 I I The procedure given in ISO 5725-2, because the
I analysis is carried out separately for each level of the
I I I
Y!, Y/1 Yi2 Yi3 Yi4 test (material), is, in fact, a two-factor fully-nested ex-
perimental design and produces two standard devi-
ations, the repeatability and reproducibility standard
Figure 2 Schematic layout of a four-factor
deviations. Factor O is the laboratory and factor 1 the
staggered-nested experiment
replication. If this design were increased by one fac-
tor, by having two operatocs in each laborato~ each
obtaining two test results under repeatability con-
The three-factor staggered-nested experiment re- ditions, then, in addition to the repeatability and
quires each laboratory i to obtain three test results. reproducibility standard deviations, one could deter-
Test results yil and yiz shall be obtained under re- mine the operator-different intermediate precision
peatability conditions, and yi~ under intermediate pre- standard deviation. Alternatively, if each laboratory
cision conditions with M factor(s) different used only one operator but repeated the experiment
(M= 1,2 or 3), for example under time-different in- on another day, the time-different intermediate pre-
termediate precision conditions (by obtaining yi~ on a cision standard deviation would be determined by this
different day from that on which yil and Yiz were ob- three-factor fully-nested experiment. The addition of
tained). a further factor to the experiment, by each laboratory
having two operators each carrying out two
In a four-factor staggered-nested experiment, Yio shall measurements and the whole experiment being re-
be obtained under intermediate precision conditions peated the next day, would allow determination of the
with one more factor different, for example, under repeatability, reproducibility, operator-different, time-
[time + operator]-different intermediate precision different, and [time + operator]-different standard
conditions by changing the day and the operator. deviations.

Again, analysis of the results of an n-factor


staggered-nested experiment is carried out by the 9.8 Comparison of fully-nested and
statistical technique analysis of variance (ANOVA) staggered-nested experimental designs
separately for each level of the test, and is described
in detail in annex C. An n-factor fully-nested experiment requires 2 - test
~esults from each laboratory, which can be an ex-
9.6 Allocation of factors in a nested cessive requirement on the%laboratories. This is the
experimental design main argument for the staggered-nested experimental
design. This design requires less test results to pro-
The allocation of the factors in a nested experimental duce the same number of standard deviations, al-
design is arranged so that the factors affected most though the analysis is slightly more complex and there
by systematic effects should be in the highest ranks is a larger uncertainty in the estimates of the standard
(O, 1, ...). and those affected most by random effects deviations due to the smaller number of test results.

9
IS 15393 (Part 3) :2003
ISO 5725-3: 1994

Annex A
(normative)

Symbols and abbreviations used in ISO 5725

u Intercept in the relationship k Mandels within-laboratory consistency test sta-


tistic
s=ai-bm
LCL Lower control limit (either action limit or warning
A Factor used to calculate the uncer-
limit)
tainty of an estimate
m General mean of the test property; level
b Slope in the relationship

s=a+bm M Number of factors considered in intermediate


precision conditions
B Component in a test result rep-
resenting the deviation of a labora- N Number of iterations
tory from the general average
n Number of test results obtained in one labora-
(laboratory component of bias)
tory at one level (i.e. per cell)
B. Component of B representing all
P Number of laboratories participating in the inter-
factors that do not change in inter-
Iaboratory experiment
mediate precision conditions
P Probability
ql), ~(2)1 etc. Components of B representing fac-
tors that vary in intermediate pre- Number of levels of the test property in the
q
cision conditions interlaboratory experiment

c Intercept in the relationship r Repeatability limit


Igs=c+dlgm
R Reproducibility limit
c, c, c Test statistics
RM Reference material
cc
crm, ,,lt, C~~i~ Critical values for statistical tests
s Estimate of a standard deviation
CDP Critical difference for probability P
: Predicted standard deviation
CRP Critical range for probability P

d Slope in the relationship T Total or sum of some expression

Igs=c+dlgm
t Number of test objects or groups
e Component in a test result rep-
resenting the random error occur- UCL Upper control limit (either action limit or warning
ring in every test result limit)

f Critical range factor w Weighting factor used in calculating a weighted


regression
FP(v1, V~) p-quantile of the F-distribution with
VI and V2 degrees of freedom
w Range of a set of test results
G Grubbs test statistic
~ Datum used for Grubbs test
h Mandels between-laboratory con-
sistency test statistic Y Test result

10
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Arithmetic mean of test results Symbols used as subscripts

Grand mean of test results c Calibration-different

Significance level E Equipment-different

Type II error probability i Identifier for a particular laboratory

Ratio of the reproducibility standard deviation to 1( ) Identifier for intermediate measures of


the repeatability standard deviation (u~/a,) precision; in brackets, identification of
the type of intermediate situation
Laboratory bias
j Identifier for a particular level
Estimate of A (ISO 5725-2).
Identifier for a group of tests or for a
Bias of the measurement method
factor (ISO 5725-3)
Estimate of 6
k Identifier for a particular test result in a
Detectable difference between two laboratory Iaboratoy i at level j
biases or the biases of two measurement
L Between-laboratory (interlaboratory)
methods
m Identifier for detectable bias
True value or accepted reference value of a test
property M Between-test-sample

Number of degrees of freedom o Operator-different

Detectable ratio between the repeatability stan- P Probability


dard deviations of method B and method A
r Repeatability
True value of a standard deviation
R Reproducibility
Component in a test result representing the
variation due to time since last calibration T Time-different

Detectable ratio between the square roots of w Within-1 aboratory (intralaboratory)


the between-laboratory mean squares of
method B and method A 1, 2, 3... For test results, numbering in the order
of obtaining them
p-quantile of the ~2-distribution with v degrees
of freedom (1), (2), (3) For test results, numbering in the order
of increasing magnitude

11
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Annex B
(normative)

Analysis of variance for fully-nested experiments

The analysis of variance described in this annex has to be carried out separately for each level of the test included
in the interlaboratory experiment. For simplicity, a subscript indicating the level of the test has not been suffixed
to the data. It should be noted that the subscript j k used in this part of ISO 5725 for factor 1 (factor O being the
laboratory), while in the other parts of ISO 5725 it is used for the level of the test.

The methods described in subclause 7.3 of ISO 5725-2:1994 should be applied to check the data for consistency
and outliers. With the designs described in this annex, the exact analysis of the data is very complicated when
some of the test results from a laboratory are missing. If it is decided that some of the test results from a lab-
oratory are stragglers or outliers and should be excluded from the analysis, then it is recommended that all the
data from that laboratory (at the level affected) should be excluded from the analysis.

B.1 Three-factor fully-nested experiment

The data obtained in the experiment are denoted by ytik, and the mean values and ranges are as follows:

jij = + (yijl
+ Yij2) ~j = ~ (Jil + fi*)

ij( 1)= lYijl Yij21 i(2) = l~il 121

where p is the number of laboratories which have participated in the interlaboratory experiment.

The total sum of squares, SST, can be subdivided as

SST = ~ ~ ~(yijk ~)z = SS0 + SS1 + SSe


i jk

where

Since the degrees of freedom for the sums of squares SS0, SS1 and SSe are p 1, p and 2P, respectively, the
ANOVA table is composed as shown in table B.1.

12
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Table 6.1 ANOVA table for a three-factor fully-nested experiment

Sum of Degreesof
Source Mean square Expected mean square
squares freedom

o Sso p1 MSO=SSO/(p -1) u; + 2af1)4 4&


1 Ssl P MS1 = SS1/p u; + 2U:1)
Residual SSe 2p MSe = SSe/(2p) a:

Total SST 4p1

The unbiased estimates s;), s~l) and s: of o~o),


,,. a~l), and o?, respectively, can be obtained from the mean squares
MSO, MSI and MSe as ~

St, = ~1 (MSO MS1)

S~I) = ~ (MS1 MSe)

s: = MSe

The estimates of the repeatability variance, one-factor-different intermediate precision variance and reproducibility
variance are, respectively, as follows:

s:

S;l) = s: i- s~l)

s; = s: + $, + St)

B.2 Four-factor fully-nested experiment

The data obtained in the experiment are denoted by ytiu, and mean values and ranges are as follows:

~jk = ~ bijkl + yijk2) ijk(l) = lYijkl Yijk21

Yi = ~ (Xl+ %2) j(3) = Iyil yjzl

where p is the number of laboratories which have participated in the interlaboratory experiment.

The total sum of squares, SST, can be subdivided as follows:

SST = ~ ~ ~, ~,~ijkl - ~)2 = SS0 + SS1 + SS2 + SSe


i Jkl

where

13
Since the degrees of freedom for the sums of squares SS0, SS1, SS2 and SSe are p 1, p, 2p and 4p, respectively,
the ANOVA table is composed as shown in table B.2.

Table B.2 ANOVA table for a four-factor fullv-nested experiment


.
Sum of Degrees of
Source Mean square Expected mean square
squares freedom

o Sso p1 MSO = SSO/(jJ 1) u: + 2uf2) + 4U~1) + 80;o)


1 Ssl P MS1 = SS1/p a; + 2U:2)+ 4U:1)
2 SS2 2p MS2 = Ss2/(2p) a: + 2uf2)
2
Residual SSe 4p MSe = SSe/(4p) Clr

Total SST 8p1

The unbiased estimates $), s~l), s;, and s: of a~o), a~l), a~2) and a:, respectively, can be obtained from the mean
squares MSO, MSI, MS2 and MSe as follows:

S~O)= ~ (MSO MS1)

s;,, = ~4 (MS2 MS1)

S~Z)= ~ (MS I MSe)

s? = MSe

The estimates of the repeatability variance, one-factor-different intermediate precision variance, two-factors dif-
ferent intermediate precision variance and reproducibility variance are, respectively, as follows:

s:

S;l) = s: + S72)

S;z) = s; + S72)+ S:l)

14
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Annex C
(normative)

Analysis of variance for staggered-nested experiments

The analysis of variance described in this annex has to be carried out separately for each level of the test included
in the interlaboratory experiment. For simplicity, a subscript indicating the level of the test has not been suffixed
to the data. It should be noted that the subscript j is used in this part of ISO 5725 for the replications within the
laboratory, while in the other parts of ISO 5725 it is used for the level of the test.

The methods described in subclause 7.3 of ISO 5725-2:1994 should be applied to check the data for consistency
and outliers. With the designs described in this annex, the exact analysis of the data is very complicated when
some of the test results from a Iaboratow are missing. If it is decided that some of the test results from a lab-
oratory are stragglers or outliers and should be excluded from the analysis, then it is recommended that all the
data from that laborat~ (at the level affected) should .be excluded from the analysis.

C.1 Three-factor staggered-nested experiment

The data obtained in the experiment within laborato~ i are denoted by yij (j= 1, 2, 3), and the mean values and
ranges are as follows:

Yi(l) = + (Yil + Yi2) i(l) = lYil - Yi21

Yi(2) = ~ (Yil + Yi2 + Y13) i(2) = lYi(l


)- Yi31

where p is the number of laboratories which have participated in the interlaboratory experiment

The total sum of squares, SST, can be subdivided as follows:

SST = ~, ~4(yo j)* = SS0 +-SS1 + SSe


i j

where

Sso = 3X (7i(*))2 3P(;)2


i

Since the degrees of freedom for the sums of squares SS0, SS1 and SSe are p 1,p and p, respectively, the
ANOVA table is composed as shown in table C.1.

15
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Table C.1 ANOVA table for a three-factor staggered-nested experiment

Source
Sum of
squares
Degreesof
freedom I Mean square
I Expected mean square
I

o Sso p1 sso/(p -1) u; + : u~,, + 30;0)

1 Ssl P Ssl /p u; + + Uf,,

Residual SSe P SSe/p 0:

Total I SST 3p1 I I I

The unbiased estimates s~o),s~l) and s; of o~o)t CJ~I) and d, respectively, can be obtained from the mean squares
MSO, MS1 and MSe as follows:

s~o)= ~ MSO ~ 5 MS1 + ~ MSe

s~l) = : MS1 ~ 3 MSe

S: = MSe

The estimates of the repeatability variance, one-factor-different intermediate precision variance and reproducibility
variance are, respectively, as follows:

s:

+;l) = s; + Sfl)

s; = s; + s~l) + s~o)

C.2 Four-factor staggered-nested experiment

The data obtained in the experiment within laboratory i are denoted by yu (j= 1,2,3, 4), and the mean values and
ranges are as follows:

X(l) = ~ (Yil + Yi2) i(1) = IYil Yi21

1(2) = ~ bil +X2+ Y~) i(2) = E(l)


Y131

Y(3) = + (Yil + YIZ? +%3 + Yi4) i(3) = 1.Ii(2) Yi41

where p is the number of taboratories which have participated in the interlaboratory experiment.

The ANOVA table is composed as shown in table C.2.

16
IS .15393 (Part 3) :2003
ISO 5725-3:1994

Table C.2 ANOVA table for a four-factor staggered-nested experiment


Degrees of

r
Sum of squares Mean square Expected mean square
freedom 1
F
o p-1 sso/(p -1) d: + ~ u~z)+ ~Eiza(l)+ 4ufoJ
2
1 Ssl /p Tzsz
P u; +U(2)+T u(,)
6
2 P ss2/p LIZ
a; + U(2)
3
Residual P SSe/p u;

I
Iu!-_ 4p1
I

C.3 Five-factor staggered-nested experiment

The data obtained in the experiment within laboratory i are denoted by yti (i= 1,2,3,4, 5), and the mean values
and ranges are as follows:

Ji(l)
= + (Yil + Yi2) i(l) = lYil Yi21

X(2) = ~ (Yil + Yi2 + Y13) i(2) = lYi(l) Yi31

~i(3) = ~ (Yil + Yi2 + Yi3 + Yi4) i(3) = IX(2) Yi41

3(4) = ~ (Yil + Yi2 + Yi3 + Yi4 + Yi5) i(4) = IX(3) Yi!jl

where p is the number of laboratories which have participated in the interlaboratory experiment

The ANOVA table is composed as shown in table C.3.

Table C.3 ANOVA table for a five-factor stacmered-nestedex~eriment


Degrees of
SourGe Sum of squares Mean square Expected mean square
freedom

D 5UX(4))2 - 5P&2 p-1 sso/(p -1)


i

1 $ ;W;4) P Ssl /p

2 + ;W;3) P ss2/p
1
3 P ss3/p
* ;42)

Residual
4 D& P SSe/p

Total -z
;,
XYij- ;)2 5p-1
I I

17
IS 15393 (Part 3) :2003
ISO 5725-3:1994

C.4 Six-faetor staggered-nested experiment

The data obtained in the experiment within Iaboratow i are denoted by yij (j= 1, 2, 3, 4, 5, 6), and the mean values
and ranges are as follows:

1(1) = ~ (Yil + Yi2) i(l)= bil Yi21

7i(2) = ~ (Yil + Yi2 + YKJ i(2) = IYi(l) Yi31

Yi(3) = ~ (Yil + Yi2 + Yi3 + Yi4) i(3) = Fi(2) .Yi41

E(4) = ; (Yil + Yi2 + Yi3 + Yi4 + Yi5) i(4) = 17i(3) Yi51

X(5) = ~ (yil + yi2 + Yi3 + yi4 + Yis + yi6) i(s) = Iyi (4) yi61

;=+
Ji(5)
z
i

where p is the number of laboratories which have participated in the interiaboratory experiment.

The ANOVA table is composed as shown in table C.4.

Table C.4 ANOVA table for a six-factor staggered-nested


-- ex~eriment
Degrees of
Source Sum of squares Mean square Expected mean square
freedom

Residual

Total

18
IS 15393 (Part 3):2003
ISO 5725-3:1994

Annex D
(informative)

Examples of the statistical analysis of intermediate precision experiments

D.1 Example 1 Obtaining the [time + D.1..2 Analysis


operator]-different intermediate precision The data yjl, y,z and Wj = ]yjl y,zl are shown in
standard deviation, q (To)#within a Specific table D.1. The analysis follows the procedure given in
laboratory at a pmticular level of the test 8.2.

A plot of the data [deviations from the mean of the


measurements on both days (yjk T) versus the
D.1.1 Background
sample number~~ is shown in figure D.1. This plot and
the application of Cochrans test indicate that the
a) Measurement method: Determination of the
ranges for samples with numbers 20 and 24 are
carbon content in steel by vacuum emission
outliers. There is a large discrepancy between the
spectrometry with test results expressed as per-
daily measurements on these samples which is most
centages by mass.
likely due to errors in recording the data. The values
for these two samples were removed from the com-
b) Source: Routine report of a steel works in Nov-
putation of the [time + operator]-different intermedi-
ember 1984.
ate precision standard deviation, ~\(T~\, which is
c) Experimental design: In a specific Iaboratoty, a calculated according to equation (12) as

r
randomly selected sample out of the materials
27
analysed on one day was analysed again on the 1
next day by a different analyst. In a month 29 pairs Sl(-ro)= WJ2 = 2,87 X 10-3
2x27 x
j=l
of such data were obtained (see table D.1 ).

Table D.1 Original data Carbon content, Yo (rdm)

Sample Sample
No.
First day Next day
I Range
No.
first day Next day Range

j Yil Yr2 I Wj j Yjl Y,2 Wj

1 0,130 0,127 0,003 16 0,149 0,144 0,005


2 0,140 0,132 0,008 17 0,044 0,044 0,000
3 0,078 0,080 0,002 18 0,127 0,122 0,005
4 0,110 0,113 0,003 19 0,050 0,048 0,002
5 0,126 0,128 0,002 20 0,042 0,146 0,104
6 0,036 0,032 0,004 21 0,150 0,145 0,005
7 0,050 0,047 0,003 22 0,135 0,133 0,002
8 0,143 0,140 0,003 23 0,044 0,045 0,001
9 0,091 0,089 0,002 24 0,100 0,161 0,061
10 0,040 0,030 0,010 25 0,132 0,131 0,001
11 0,110 0,113 0,003 26 0,047 0,045 0,002
12 0,142 0,145 0,003 27 0,168 0,165 0,003
13 0,143 0,150 0,007 28 0,092 0,088 0,004
14 0,169 0,165 0,004 29 0,041 0,043 0,002
I
15 0,169 0,173 0,004

19
IS 15393 (Part 3) :2003
ISO 5725-3:1994

0,06

0,02

c1
1
++- .nnfJ~ ln IJn
r In IJn+
0 a[ 1++++ u ++~ +++0
+ + ++
+

-0,02

-0,04

L
First day
-1- Next day
r1

-0,06
I I I I I I I I I I I I I I I I
1 5 10 15 20 25 30
Sample number, j _

Figure D.1 Carbon content in steel Deviations from the mean of the measurements on both days
versus the sample number

D.2 Example 2 Obtaining the b) Source lSO/TC 17, Stee//SC 1, Methods of de-
time-different intermediate precision termination of chemica/ composition. Experiment
carried out in May 1985.
standard deviation by interlaboratory
experiment
c) Experimental design: A three-factor staggered-
nested experiment was carried out with 20 lab-
D.2.I Background oratories each reporting two test results obtained
under repeatability conditions on one day, fol-
a) Measurement method: Determination of the lowed by one further test result on the next day
vanadium content in steel by the atomic absorp- at each of the six levels included in the exper-
tion spectrometric method descriied in the in- iment. All measurements in any laboratory were
structions for the experiment. Test results are carried out by one operator, using the same
expressed as percentages by mass. measurement equipment.

20
IS 15393 (Part 3) :2003
ISO 5725-3:1994

D.2.2 Analysis oratory. There is a large discrepancy between the test


result for day 2 and the mean value for day 1, which
The data for all six levels are given in table D.2. is ve~ large compared with the test results for the
other laboratories. This laboratory was removed from
The analysis of variance is presented for only one of the calculations of the precision measures.
the levels, i.e. level 1.

A plot of the data (test results for day 1 and day 2 In accordance with C.1 of annex C, Wi(l), Wi(z) and
versus laboratory number i) is shown in figure D.2. ](p) were computed and the results are shown in ta-
This plot indicates that laboratory 20 is an outlier lab- bieD.3.

150

I
-Y
0
l Day 1, first test result
-1- Day 1, second test result
x Iko
* clay 2
~
.5 I I
aQ
E
2 130
c
0

:
I I
c

~ 120
+ *

*
1 1
110 ,+

1

n+


+CI*
100
c1
+
+
*m
*
f -r
90 -m
1

*
80

I I 1 I I I I I I I I 1
70
1 4 8 ?2 16 20
Laboratory number, i

Figure D.2 Vanadium content in steel Test results for day 1 and 2 at level 1 versus laboratory number

21
Level 1 (0,01%) Level 2 (0,04 %) Level 3 (O,1 %) Level 4 (0,2 %) Level 5 (0,5 %) Level 6 (O,75 7.)
Laboratory
No. Day 1 Day 2 Day 1 Day 2 Day 1 Day 2 Day 1 Day 2 Day 1 Day 2 Day 1 Day 2

i Yil Y(2 Y(3 Yil Y,2 Yi3 Yil Y,2 Y,3 Y,l Y,2 Ya Yil Y,3 Yi2 Y,l Yi2 Yi3

1 0,0091 0,0102 0,0098 0,0382 0,b38 8 0,0385 0,101 0,103 0,102 0,214 0,211 0,210 0,514 0,510 0,513 0,755 0,753 0<751

2 0,0100 0,0100 0,0090 0,0410 0,041 0 0,0390 0,111 0,111 0,108 0,220 0,220 0,215 0,520 0,540 0,540 0,800 0,755 0,750

3 0,0095 0,0090 0,0094 0,0390 0,0380 0,0370 0,108 0,110 0,107 0,213 0,215 0,215 0,500 0,514 0,504 0,738 0,730 0,724

4 0,0080 0,0083 0,0077 0,0374 0,0361 0,0382 0,109 0,106 0,104 0,214 0,222 0,201 0,519 0,518 0,518 0,744 0,742 0,732

5 0,0100 0,0100 0,0100 0,0350 0,0370 0,0370 0,103 0,103 0,110 0,210 0,210 0,205 0,495 0,500 0,512 0,743 0,753 0,750

6 0,0089 0,0094 0,0094 0,0368 0,0368 0,0377 0,106 0,106 0,108 0,232 0,240 0,221 0,526 0,532 0,513 0,733 0,740 0,746
7 0,0098 0,0099 0,0101 0,0376 0,0380 0,0384 0,107 0,105 0,108 0,215 0,215 0,216 0,521 0,519 0,526 0,754 0,756 0,756

8 0,0096 0,0094 O,odg 9 0,0379 0,0366 0,0379 0,108 0,107 0,108 0,193 0,195 0,210 0,507 0,493 0,511 0,732 0,729 0,732

9 0,0104 0,0094 0,0100 0,0365 0,0370 0,0367 0,104 0,106 0,105 0,211 0,205 0,213 0,509 0,515 0,515 0,734 0,738 0.747
10 0,0107 0,0118 0,0097 0,0370 0,0375 0,0380 0,105 0,110 0,105 0,210 0,220 0,225 tJ,520 0,520 0,525 0,760 0,760 0,765

11 0,0100 0,0100 0,0090 0,0380 0,0380 0,0375 0,102 0,102 0,102 0,213 0,211 0,214 0,513 0,516 0,514 0,746 0,748 0,746
12 0,0109 0,0115 0,0097 0,0390 0,0390 0,0390 0,101 0,108 0,105 0,208 0,215 0,210 0,509 0,528 0,510 0,758 0,748 0,750

13 0,0100 0,0095 0,0095 0,0375 0,0375 0,0375 0,103 0,104 0,108 0,212 0,222 0,215 0,510 0,520 0,505 0,735 0,755 0,750

14 0,0096 0,0096 0,0100 0,0374 0,0374 0,0389 0,104 0,106 0,110 0,218 0;218 0,212 0,520 0,528 0,522 0,740 0,735 0,742

15 0,0099 0,0091 0,0082 0,0381 0,0375 0,0392 0,109 0,106 0,107 0,214 o,2io 0,211 0,510 0,510 0,515 0,749 0,729 0,744

16 0,0098 0,0100 0,0095 0,0373 0,0377 0,0397 0,105 0,105 0,104 0,215 0,212 0,218 0,519 0,517 0,531 0,754 0,751 0,759

17 0,0105 0,0102 0,0112 0,0389 0,0382 0,0373 0,107 0,108 0,104 0,214 0,210 0,209 0,517 0,515 0,514 0,735 0,728 0,741

18 0,0103 0,0105 0,0118 0,0382 0,0380 0,0374 0,103 0,104 0,103 0,224 0,218 0,217 0,515 0,514 0,517 0,788 0,798 0,787

19 0,0098 0,0096 0,0104 0,0383 0,0375 0,0366 0,110 0,109 0,104 0,217 0,215 0,215 0,530 0,525 0,520 0,755 0,745 0,740

. 20 0,0140 0,0140 0,0100 0,0370 0,0408 0,0369 0,104 0,106 0,107 0,214 0,214 0,203 0,518 0,518 0,481 0,730 0,737 0,658
IS 15393 (Part 3) :2003
ISO 5725-3:1994

1
X(2) = 0,00979825
Table D.3 Values of w,(,), w,~,,and Y,,,, Y=19~
,(-)
.,!, .,-,
-

Laboratory
No. Wi(l) i(2) X(2) From these values, the sums of squares SS0, SS1 and
i SSe are obtained, and the ANOVA table is prepared
1 0,!)01 1 0,00015 0,009700 as shown in table D.4.
2 0,0000 0,00100 0,009667
3 0,0005 0,00015 0,009300 The unbiased estimates of variances between labora-
4 0,0003 0,00045 0,008000 tories S$), between days within a laboratory s~l), and
5 0,0000 0,00000 0,010000 the estimated repeatability variance s: are obtained
6 0,0005 0,00025 0,009233 as
7 0,0001 0,00025 0,009933
8 0,0002 0,00040 0,009633 S$, = 0,278 X 10-s
9 0,0010 0,00010 0,009933
10 0,001 1 0,001 55 0,010733 $, = 0,218 X 10-G
11 0,0000 0,00100 0,009667
12 0,0006 0,001 50 0,010700
13
S: =0,14.5X 10-6
0,0005 0,00025 0,009667
14 0,0000 0,00040 0,009733
15 0,0008 0,00130 0,009067
16 0,0002 0,00040 0,009767 The reproducibility standard deviation SR, time-
17 0,0003 0,00085 0,010633 different intermediate precision standard deviation
18 0,0002 0,00140 0,010867 S(F), and repeatability standard deviation s, are ob-
19 0,000 2 0,00070 0,009933 tained as

The sums of squares of Wi(l), Wi(z) and Y(2) and the


SR =
d t + fI) + $) = 0,801 X 10-3
mean value ~ are computed as

z i
Wfi,, 5,52x10-6

E
w,~2)
12,44x 10-6
Sr =
J_ S2
r = 0,381 X 10-3

The values of these standard deviations for the six


X(X,2,)2= 1 832,16X ,0-6 levels of vanadium content are summarized in
i table D.5 and shown in figure D.3.

Table D,4 ANOVA table Vanadium content

Sum of Degreas of Mean


Source Expected mean square
squares freedom square

O (laboratory) 24,16 x10- 18 1,342 x10-G a; + : u;,, + 3fJ&,

1 (day) 8,29 10-6 42


X 19 0,436 X 10-6 u; + ~u(l)

Residual 2,76 X 10-6 19 0,145 x 10-6 a;


t ,
Total I 35,21 X10-6 I 56 I I I

23
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Table D.5 -Values of s,, Sin,ands~


.,. , for six levels of vanadium content in steel

Outlier
1
Level laboratories Average (%) s, (%) sl(T) @) s~ (%)
No.

1 20 0,0098 0,381 X 10-3 0,603X 10-3 0,801 X 10-3


2 2 0,0378 0,820 X 10-3 0,902X 10-3 0,954 x 10-3
3 0,1059 1,739 X1 O-3 2,305X 10-3 2,650 X 10-3

4 6 and 8 0,2138 3,524 X 10-3 4,7 IO X1 O-3 4,826 X 10-3

5 20 0,5164 6,237 X 10-3 6,436 X 10-3 9,412 x10-3


6 20 0,7484 9,545 x 10-3 9,545 x 10-3 15,962 X 10-3

12

.- ..
.-
-----

o
0,001 0.01 0,1
Concentration level, %

Figure D.3 Vanadium content in steel Repeatability standard deviation s,, time-different intermh@ate
precision standard deviation Slfl)and reproducibility standard deviations~ as functions of the concentra~n
level

24
IS 15393 (Part 3) :2003
ISO 5725-3:1994

Annex E
(informative)

Bibliography

[1] ISO 3534-2:1993, Statistics Vocabulary and Part 5: Alternative methods for the determi-
symbols Part 2: Statistical quality control. nation of the precision of a standard measure-
ment method.
[2] ISO 3534-3:1985, Statistics Vocabulary and
symbols Part 3: Design of experiments. [5] ISO 5725-6:1994, Accuracy (trueness and pre-
cision) of measurement methods and results
[3] ISO 5725-4:1994, Accuracy (trueness and pre- Part 6: Use in practice of accuracy values.
cision) of measurement methods and resu/ts
Part 4: Basic methods for the determination of [6] WINER, B.J. Statistical principles in experimental
the trueness of a standard measurement design. McGraw-Hill, 1962.
method,
[7] SNEDECOR,G.W. and COCHRAN,W.G. Statistical
[4] ISO 5725-5:11, Accuracy (trueness and pre- methods. Iowa University Press, 1967.
cision) of measurement methods and results

1) To be published.

25
( Continuedfrom secomfcover)

IS No. Title

IS 15393 ( Part 4 ) : 2003/ Accuracy ( trueness and precision ) of measurement methods and
ISO 5725-4:1994 results : Part 4 Basic methods for the determination of the trueness of
a standard measurement method

IS 15393 ( Part 5 ) : 2003/ Accuracy ( trueness and precision ) of measurement methods and
ISO 5725-5:1994 results: Part 5 Alternative methods for the determination of the precision
of a standard measurement method

IS 15393 ( Part 6 ) : 2003/ Accuracy ( trueness and precision ) of measurement methods and
ISO 5725-6:1994 results : Part 6 Use in practice of accuracy values

The concerned Technical Committee has reviewed the provisions of following International Standards
referred in this adopted standard and decided that it is acceptable for use with this standard:

1s0 3534-1:1993 Statistics Vocabulary and symbols Part 1: Probability and general
statistical terms

ISO Guide 33:1989 Uses of certified reference materials

ISO Guide 35:1989 Certification of reference materials General and statistical principles

This standard also gives Bibliography in Annex E, -which is informative.


Bureau of Indian Standards

BIS is a statutory institution established under the Bureau of Indian Standards Act, 1986 to promote
harmonious development of the activities of standardization, marking and quality certification of goods and
attending to connected matters in the country,

Copyright

BIS has the copyright of all its publications. No part of these publications maybe reproduced in any form without
the prior permission in writing of BIS. This does not preclude the free use, in the course of implementing the
standard, of necessary details, such as symbols and sizes, type or grade designations. Enquiries relating to
copyright be addressed to the Director (Publications), BIS.

Review of Indian Standards

Amendments are issued to standards as the need arises on the basis of comments. Standards are also reviewed
periodically; a standard along with amendments is reaffirmed when such review indicates that no changes are
needed; if the review indicates that changes are needed, it is taken up for revision. Users of Indian Standards
should ascertain that they are in possession of the latest amendments or edition by referring to the latest issue
of BIS Catalogue and Standards : Monthly Additions.

This Indian Standard has been developed from Doc : No. BP 25 ( 0190 ).

Amendments Issued Since Pu-blication

Amend No. Date of Issue Text Affected

BUREAU OF INDIAN STANDARDS


Headquarters:
Manak Bhavan, 9 Bahadur Shah Zafar Marg, New Delhi 110002 Telegrams: Manaksanstha
Telephones: 23230131,23233375,2323 9402 ( Common to all offices)
Regional Offices: Telephone
Central: Manak Bhavan, 9 Bahadur Shah Zafar Marg 23237617
NEW DELHI 110002 { 23233841

Eastern: 1/14 C. I. T. Scheme VII M, V. I. P. Road, Kankurgachi 23378499,23378561


KOLKATA 700054 { 23378626,23379120
Northern: SCO 335-336, Sector 34-A: CHANDIGARH 160022 603843
{ 609285
Southern: C. 1.T. Campus, IV Cross Road, CHENNAI 600113 22541216,22541442
{ 22542519,22542315
Western : Manakalaya, E9 MIDC, Marol, Andheri (East) 28329295,28327858
MUMBAI 400093 { 28327891,28327892

Branches : AHMEDABAD. BANGALORE. BHOPAL. BHUBANESHWAR. COIMBATORE.


FARIDABAD. GHAZIABAD. GUWAHATI. HYDERABAD. JAIPUR. KANPUR.
LUCKNOW. NAGPUR. NALAGARH. PATNA. PUNE. RAJKOT. THIRUVANANTHAPW.
VISAKHAPATNAM.
Printed at New India Printing Press, Khurja, India