Documente Academic
Documente Profesional
Documente Cultură
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/247838253
CITATIONS READS
2 1,776
1 author:
SEE PROFILE
All content following this page was uploaded by Elart von Collani on 11 November 2014.
It turns out that the purpose of ISO 2859-1 is obsolete, that the formulation is ambiguous,
that it does not represent the state of the art and nally that it hinders necessary future
technological development in the eld.
Primarily, those characteristics shall be included which are suitable for worldwide
(universal) acceptance. Where necessary, owing to dierences in legislation, climate,
environment, economies, social conditions, trade patterns, etc., several opinions may
be indicated.
In the following, one of the most popular statistical standards shall be assessed by means
of the general requirements and rules cited above. These rules have proved to be extremely
successful in all areas of technology and have emerged as one pathnder for globalization.
Applying the above directives to a statistical standard would demand the following: First,
the purpose has to be specied and has to be translated into statistical terms, next, the
required statistical performance criteria should be stated and nally, some rules must be
given outlining statistical procedures which guarantee that the purpose is achieved but
which do not hinder technological advancements.
2 Purpose of a Standard
The scope of a standard refers to the purpose and possibly limitations with respect to
situations when the standard can be applied. As mentioned above, the purpose is of
utmost importance and, therefore, should be given very clearly. Furthermore, it should
be assured that it coincides with specied needs of potential users.
Any limitation for possible application should be necessary and given in an unambiguous
way, as otherwise the procedure is applied in non admitted situations which would lead
to wrong results.
Accordingly, the standard assumes a situation with two parties (supplier and consumer)
and can be used by the consumer as a means for generating pressure on the supplier and
limiting the risk for accepting lots of poor quality. Exerting economic and psychological
pressure is stated as the primary aim of the standard. For achieving this aim, one would
expect rather psychological than statistical expert knowledge. Nevertheless, the standard
is classied as a statistical one.
The rst question to be considered refers to the need of methods for exerting pressure
on suppliers with respect to the facilitation of international trade and communication.
In view of global competition and in view of movements like Six Sigma and the im-
plementation of continuous quality improvement strategies, one can hardly think of any
non-military consumer needing special tools for generating pressure on a supplier. Any
supplier aording poor quality processes is almost immediately taken from the global mar-
ket with or without economic pressure of its consumers by means of a sampling scheme.
Thus, we may conclude that the primary purpose of ISO 2859-1 is of no relevance for fa-
cilitating international trade and communication as it assumes an isolated situation
where the consumer is more or less at the mercy of one supplier and, therefore, needs
some means to exert pressure on the supplier.
The second purpose mentioned refers to an upper limit for the risk of accepting the
occasional poor lot. At a rst glance, the second purpose seems to be closer to existing
needs. But for understanding its meaning, one should know the meaning of risk, upper
limit, occasional and, of course, poor lot. However, no explanations can be found in
the standard. The Index of terms used in this part of ISO 2859 does not contain any
of the terms needed to understand the second purpose properly.
In [9] two reasons for applying acceptance sampling in modern industrial environments
are identied. These are legal liability requirements on the one hand and monitoring the
actual quality level on the other. Neither of them is mentioned in ISO 2859-1.
Sampling schemes designated in this part of ISO 2859 are applicable, but
not limited, to inspection of a) end items; b) components and raw materials;
c) operations; d) materials in process; e) supplies in storage; f ) maintenance
operations; g) data or records; h) administrative procedures.
These schemes are intended primarily to be used for a continuing series of lots,
that is, a series long enough to allow the switching rules (9.3) to be applied.
These rules provide:
268 Review of Statistical Standards and Specications
Having in mind the primary purpose of ISO 2859-1 to exert pressure assuming a situation
with two opposing parties, one should have expected that situations with respect to the
two parties (supplier and consumer) are listed in order to illustrate possible elds of
application. Listing possible items to be tested must be confusing as in most cases the
two parties cannot be clearly identied and how the dierent situations with respect to
the nature of items can be incorporated into the basic assumptions.
Next, it is stated that these schemes are primarily to be used for a continuous series of
lots, but a few lines further, the reader is confronted with sampling plans in this part of
ISO 2859 may also be used for the inspection of lots in isolation. It remains unclear how
long a series of lots must be in order not to be looked upon as isolated lots. Moreover,
there is no case dierentiation between series of lots delivered by one or several suppliers.
Neither of these aims is quantied. It is not stated how the pressure on the supplier
is measured and there is no denition of consumers risk, which, of course, must be
Review of Statistical Standards and Specications 269
It appears that the authors of ISO 2859-1 have made an attempt to come up with some
absolute (philosophical) denitions detached from the actual problem to explain the terms
needed for introducing (simple) sampling plans. Having in mind the aim to explain lot
inspection and nothing else, it is hardly possible to guess the meaning of the denitions
given in ISO 2859-1. The authors must have the same feeling, as in many cases, they see
the necessity to further explaine their denitions by supplementary notes. In fact, the
meaning of most of the terms is clear without reading the denitions, but reading them
actually generates confusion.
There is another strange fact. The section aims at introducing terms and symbols. How-
ever, the terms and the corresponding symbols are persistently given separately.
The sampling plans shall support special decisions with respect to given lots. There
are three decisions possible: acceptance with continuation of sampling, rejection with
continuation of sampling and rejection with discontinuation of sampling. In order to
assure an appropriate application, the meaning of the three decisions should be stated.
Attempts are made in Section 7 Acceptance and non-acceptance and in Section 9 Nor-
mal, tightened and reduced inspection. The respective entries are as follows.
270 Review of Statistical Standards and Specications
Acceptability of lots
The term non-acceptance is used in this context for rejection when it refers
to the result of following the procedure. Forms of the term reject are retained
when they refer to actions the consumer may take, as in rejection number.
Besides rejection of lots, there is another more severe means for exerting pressure on the
supplier provided in the standard, namely discontinuation of inspection. In Section 9 we
nd the following two statements:
Apparently, in Section 9.2 not the continuation of inspection as indicated by the head-
ing, but the continuation of the dierent types of sampling is subject of the entry,
where the meaning of applying the switching rule independently to classes of noncon-
forming items is unclear.
The directive in 9.4 is unclear and confusing. It says that if ve consecutive lots are
rejected by tightened inspection then this part of 2859 (i.e., Part 1) does not anymore
apply, but, unfortunately, it is not mentioned which part applies instead. There is no hint
given, how to deal with lots delivered after the decision of discontinuation of inspection
has been made. Are these lots to be rejected or used according to the original purpose
instead of their bad quality. However, it seems to be clear that the decision to change the
supplier is not considered in ISO 2859-1. Thus, the user is left alone with an irrelevant
purpose and obscure denitions and directives.
Review of Statistical Standards and Specications 271
The procedures in ISO 2959-1 aiming at exerting pressure on the supplier are indexed
in terms of the Acceptance Quality Limit3 (AQL). Thus, the AQL plays a decisive role
and its selection heavily determines the performance of the procedure, thus necessitating
a clear denition of the AQL.
acceptance quality limit (AQL): quality level that is the worst tolerable
process average when a continuing series of lots is submitted for acceptance
sampling
Note 1 This concept only applies when a sampling scheme with rules for
switching and for discontinuation, such as in ISO 2859-1 or ISO 3951, is used.
NOTE 2 Although individual lots with quality as bad as the acceptance qual-
ity limit may be accepted with fairly high probability, the designation of an
acceptance quality limit does not suggest that this is a desirable quality level.
Sampling schemes found in International Standards such as this part of ISO
2859, with their rules for switching and for discontinuation of sampling inspec-
tion, are designed to encourage suppliers to have process averages consistently
better than the AQL. Otherwise, there is a high risk that the inspection severity
will be switched to tightened inspection under which the criteria for lot accep-
tance become more demanding. Once on tightened inspection, unless action is
taken to improve the process, it is very likely that the rule requiring discontin-
uation of sampling inspection pending such improvement will be invoked.
It is a characteristic feature of ISO 2859 that many of the denitions or explanations need
additional explanations in form of notes serving as explanations of the explanation.
The self posed general requirements for ISO standards demand among other things that
the document shall be consistent, clear and accurate. The above denition is inconsis-
tent, unclear an inaccurate. It is inconsistent as on the one hand a series of lots is required
for applying this concept (probably the AQL concept?) and on the other hand isolated
lots are explicitly admitted; it is unclear as the meaning of for the purpose of sampling
inspection is not stated; and it is inaccurate as neither the process average nor the limit
of a satisfactory process average are exemplied.
There is another attempt made in ISO 2859-1 to introduce properly the term AQL. Section
5 of ISO 2859-1 is headed Acceptance quality limit (AQL) and, therefore, one could
hope that it contains a more satisfactory denition of the fundamental concept AQL, as
otherwise a rational selection would be impossible. Unfortunately, there is no subsection
containing a denition, instead, there are three subsections headed Use and application,
3
Better known as Acceptable Quality Level.
272 Review of Statistical Standards and Specications
Specifying AQLs and Preferred AQLs. The main parts of these subsections are cited
and briey commented below.
The AQL is a parameter of the sampling scheme and should not be confused
with the process average which describes the operating level of the manufactur-
ing process. It is expected that the process average will be less than or equal to
the AQL to avoid excessive rejections under this system.
In 3.6 it is said that the AQL is the limit of a satisfactory process average, however,
in 5.1 we learn that the AQL is a designated value of percent nonconforming (or noncon-
formities per 100 units) that will be accepted most of the time by the sampling scheme to
be used. To make things even more confusing, the acceptance probability depends upon
the sample size. It appears that the AQL is a quantity related in a rather obscure way
to something called process average and to the probability of acceptance, which itself is
related to the sample size. The process average is vaguely circumscribed by an operating
level. And the notice not to interpret the AQL as a right of a supplier to deliver knowingly
nonconforming items does not at all contribute to more clarity.
In order to select an AQL value in a rational way, one should know its meaning. Unfor-
tunately, this is not possible as it is not stated clearly in the standard. Therefore, one
might hope to learn it from the next subsection.
Again, hoping was in vain. It is said that the responsible authority (= concept used
to maintain neutrality . . .) shall designate the value and that the AQL shall not exceed
10% (!) when the quality level is expressed as percent of nonconforming item (= that
which can be individually described and considered). Anyway it is comforting that the
standard admits to select in dierent situations dierent values for the AQL.
The series of values of AQLs given in the tables are known as the preferred
series of AQLs. If, for any product, an AQL is designated other than one of
these values, these tables are not applicable.
Evidently, no clarication about the meaning of the AQL is given. The tabulated values
are called preferred AQLs probably because they are tabulated, but not because they
are marked by any favorable property.
The justication of the statement that ISO 2859-1 cannot be looked upon as a statistical
method becomes obvious when looking at the withdrawn version of ISO 2859-1 from 1989,
where it is very clearly stated Section 1 Scope.
This part of ISO 2859 is not intended as a procedure for estimating lot quality
or for segregating lots.
274 Review of Statistical Standards and Specications
It is unclear why this addendum has been abandoned in the actual version. However, it
shows that ISO 2859 does not pursue any clear statistical aim, as statistical aims could
be
These aims cannot be found in the purpose of the actual version and have been explicitly
excluded from the purpose in a past version.
6 Conclusion
ISO 2859-1 (1999) has almost 100 pages with more than 65 pages of tables. There are no
requirements expressed in terms of performance, but all requirements are given by design
or descriptive characteristics in form of tables. The excessive number of tables is necessary
because the sampling scheme cannot be characterized by global (statistical) performance
properties. Each sampling plan performs dierently and, therefore, needs special tables
for illustration.
In view of the primary objective of ISO 2859-1 to exert psychological pressure on the
supplier, the question whether it hinders or advances technological development is idle, as
exerting psychological pressure is not part of technology. However, ISO 2859-1 is widely
used in industry for assessing the quality of lots and for sorting-out poor quality. This
erroneous application is abetted by the fact that in the US military and civilian precursors
of ISO 2859-1 and also in the textbooks on statistical quality control the actual purpose
was and is not mentioned. For instance, the American National Standard ANSI/ASQC
Z1.4-1981 which corresponds directly to MIL-STD-105D states the following purpose:
As the primary purpose (exerting pressure) of ISO 2859-1 has been not explicitly stated
for many decades, ISO 2859-1 has been considered erroneously as a statistical standard
aiming at assessing the quality of incoming or outgoing lots. Moreover, being an ISO
standard, it is mistaken as the state of art and recommended in courses and textbooks on
Review of Statistical Standards and Specications 275
Many of the above arguments are not new. Moreover, in a series of papers by v. Collani
[4, 5] and Gob [10, 12] some more fundamental deciencies of AQL-sampling plans were
revealed and discussed. The International Organization of Standardization has reacted
on these ndings by slightly changing the wording instead of seriously considering the
needs of industry by developing and oering statistical standards instead of military or
psychological standards.
References
[1] American National Standard (1978): Sampling Procedures and Tables for Inspection
by Attribute. American Society for Quality Control, Wisconsin.
[2] American National Standard (1981): Terms, Symbols and Denitions for Acceptance
Sampling. American Society for Quality Control, Wisconsin.
[3] Banks, J. (1989): Principles of Quality Control. John Wiley & Sons, New York.
[5] v. Collani, E. (1992): The Pitfall of Variables Sampling. In: Frontiers in Statistical
Quality Control 4, Physica Verlag, Heidelberg, 91 - 99.
[6] v. Collani, E. & K. Drager (2001): Binomial Distribution Handbook for Scientists
and Engineers. Birkhauser, Boston.
[7] v. Collani, E. (2004): Theoretical Stochastics. In: Dening the Science of Stochas-
tics. Ed. E. v. Collani, Heldermann Verlag, Lemgo, 147-174.
[8] v. Collani, E. (2004): Empirical Stochastics. In: Dening the Science of Stochastics.
Ed. E. v. Collani, Heldermann Verlag, Lemgo, 175-213.
[9] v. Collani, E. and Palcat, F. (2004): How some ISO standards complicate quality
improvement. In: Proceedings of the VIII International Workshop on ISQC04, Eds:
Grzegorzewski, P., Mrowka, E., Hryniewicz, O., Lenz, H.L., Wilrich, P.Th., Warsaw,
45-59.
[10] Gob, R. (1996): An Elementary Model of Statistical Lot Inspection and its Appli-
cation to Sampling Variables. Metrika 44, 135-163.
276 Review of Statistical Standards and Specications
[11] Gob, R. (1996): Test of Signicance for the Mean of a Finite Lot. Metrika 44,
223-238.
[12] Gob, R. (2001): Methodological Foundation of Statistical Lot Inspection. In: Fron-
tiers in Statistical Quality Control 6, Physica Verlag, Heidelberg, 3-24.
[13] Dodge, H.F. (1969): Note on the Evolution of Acceptance Sampling Plans, Part II.
Journal of Quality Technology 1, 155-162.
[14] Duncan, A.J. (1965): Quality Control and Industrial Statistics. 3rd ed., Richard D.
Irwin, Homewood.
[15] Duncan, A.J. (1972): Quality Standards. Journal of Quality Technology 4, 102-109.
[16] Juran, J.M. (1962): Quality Control Handbook. 2nd ed., McGraw-Hill, New York.
[17] Kendall, M.G. W.R. Buckland (1960): A Dictionary of Statistical Terms. 2nd ed.,
Oliver and Boyd, Edinburgh.
[18] Kotz, S. N.J. Johnson (1981): Encyclopedia of Statistical Sciences., Vol. 1, John
Wiley & Sons, New York.
[19] Schilling, E.G. (1892): Acceptance Sampling in Quality Control. Marcel Dekker,
New York.
[20] ISO 2859-1 (1989): Sampling procedures for inspection by attribute - Part 1. Inter-
national Organization for Standardization, Geneva.
[21] ISO TC 69 Application of statistical methods (2003): Business Plan 2003. Interna-
tional Organization for Standardization, Geneva.