Professional Documents
Culture Documents
Statistical Interpretation of Data - : Guide To
Statistical Interpretation of Data - : Guide To
Statistical Interpretation of Data - : Guide To
1975
ISO 3207:1975
CONFIRMED
OCTOBER 1985
Guide to
Statistical
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
interpretation of
data —
Part 3: Determination of a statistical
tolerance interval
UDC 31:519.24:312:519.243.24
BS 2846-3:1975
Co-operating organizations
Biometrika
British National Committee for Surface Active Agents
British Paper and Board Industry Federation (PIF)
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
Contents
Page
Co-operating organizations Inside front cover
Foreword ii
Section 1. Formal presentation of results
General remarks 1
Section 2. Examples
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
Introductory remarks 6
Numerical examples 7
Annex A Case of any distribution 9
Annex B Statistical tables 11
Nomograph 1 — One-sided statistical tolerance interval 16
Nomograph 2 — Two-sided statistical tolerance interval 17
Table 1 — One-sided statistical tolerance interval (known variance) 2
Table 2 — Two-sided statistical tolerance interval (known variance) 3
Table 3 — One-sided statistical tolerance interval (unknown variance) 4
Table 4 — Two-sided statistical tolerance interval (unknown variance) 5
Table 5 — One-sided statistical tolerance interval, Ö known, m unknown.
Values of the coefficient k1 (n, p, 1 – µ) 11
Table 6 — Two-sided statistical tolerance interval, Ö known, m unknown.
Values of the coefficient k ′1 (n, p, 1 – µ) 12
Table 7 — One-sided statistical tolerance interval, m, Ö unknown.
Values of the coefficient k2 (n, p, 1 – µ) 13
Table 8 — Two-sided statistical tolerance interval, m, Ö unknown.
Values of the coefficient k ′2 (n, p, 1 – µ) 14
Table 9 — Non-parametric one-sided statistical tolerance
intervals — Sample size n for a proportion p at confidence level 1 – µ 15
Table 10 — Non-parametric two-sided tolerance intervals — Sample
size n for a proportion p at confidence level 1 – µ 15
© BSI 02-2000 i
BS 2846-3:1975
Foreword
The correct interpretation and presentation of test results has been assuming
increasing importance in the analysis of data obtained from manufacturing
processes based on sample determinations and prototype evaluation in industry,
commerce and educational institutions. It was for this reason that
sub-committee 2 of Technical Committee 69, Application of Statistical Methods,
of the International Organization for Standardization (ISO), was charged with
the task of establishing a guide to statistical methods for the interpretation of test
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
ii © BSI 02-2000
BS 2846-3:1975
additional information.
Annex A provides procedures for the determination of statistical tolerance
intervals which may be utilized irrespective of the type of distribution applicable.
This Part of this British Standard is identical to ISO 3207 “Statistical
interpretation of data — Determination of a statistical tolerance interval”.
Because the text of an International Standard has been copied, it includes some
minor material not appropriate to a British Standard.
For the purposes of this British Standard the text of ISO 3207 given in this
publication should be modified as follows:
Terminology. The words “British Standard” should replace “International
Standard” wherever they appear.
The decimal point should replace the decimal comma wherever it appears.
Cross reference. The references to other International Standards should be
replaced by references to British Standards as follows.
Clause Reference to ISO Standard Appropriate British Standard
General remarks ISO/DIS 2854 “Statistical BS 2846 “Statistical
interpretation of data — interpretation of data”
Techniques of estimation and Part 43) Techniques of
tests relating to means and estimation and tests relating to
variances” means and variances”
ISO. . . .3) “Statistical BS. . . .3) “Tests for departure
interpretation of data — from normality”
Normality tests”
The following British Standards also provide practical guidance on the
application of statistical methods.
BS 600, Application of statistical methods to industrial standardization and
quality control.
BS 1313, Fraction-defective charts for quality control.
BS 2564, Control chart technique when manufacturing to a specification, with
special reference to articles machined to dimensional tolerances.
BS 6000, Guide to the use of BS 6001. Sampling procedures and tables for
inspection by attributes.
BS 6001, Sampling procedures and tables for inspection by attributes.
BS 6002, Sampling procedures and charts for inspection by variables for per cent
defective3).
BS. . . ., Determination and statistical analysis of precision data for a standard
test method3).
3) In course of preparation.
A British Standard does not purport to include all the necessary provisions of a
contract. Users of British Standards are responsible for their correct application.
Compliance with a British Standard does not of itself confer immunity
from legal obligations.
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
Summary of pages
This document comprises a front cover, an inside front cover, pages i to iv,
pages 1 to 17 and a back cover.
This standard has been updated (see copyright date) and may have had
amendments incorporated. This will be indicated in the amendment table on the
inside front cover.
iv © BSI 02-2000
BS 2846-3:1975
General remarks
1) This International Standard specifies methods enabling a sample to be used as the basis for
determining a statistical tolerance interval, i.e. an interval such that there is a fixed probability
(confidence level) that the interval will contain at least a proportion p of the population from which the
sample is taken. The statistical tolerance interval may be two-sided or one-sided. The limits of the
interval are called “statistical tolerance limits”; they are also called “natural limits of the process”.
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
2) These methods are applicable only where it may be assumed that in the population under
consideration the sample units have been selected at random and are independent.
3) The methods described below apply also only on condition that the distribution of the characteristic
being studied is normal. The requirement of normality is more important here than for the inferences
on means and differences between means in ISO 2854, Statistical interpretation of data — Techniques
of estimation and tests relating to means and variances.
4) In order to check the hypothesis of normality, the methods laid down in ISO. . . , Statistical
interpretation of data — Normality tests4), are used.
5) Where the hypothesis of normality has to be rejected or where there is some reason to doubt its
validity, one may envisage transforming the variate to make it normal or applying the method described
in the introductory remark of Annex A of this International Standard.
It is also possible to apply now methods which allow the determination of statistical tolerance intervals
for other distribution forms than normal distributions. The description of these methods has not been
considered in this International Standard.
6) In determining a statistical tolerance interval, it is desirable in connection with the origin or the
method of collection of data to give all information that may assist in their statistical analysis, in
particular the smallest unit or fraction of a measurement unit having practical significance.
7) No elimination or potential correction of individual data that are doubtful shall be carried out unless
there are experimental, technical or obvious reasons to provide circumstantial justification of such
elimination or correction.
In every instance, mention shall be made of the data eliminated or corrected.
8) As stated in 1), the confidence level 1 – µ is the probability that the statistical tolerance interval will
contain at least a proportion p of the population. The risk of this interval containing less than a
proportion p of the population is µ. The most usual values of 1 – µ are 0,95 and 0,99 (µ = 0,05 and 0,01).
This means that if statistical tolerance intervals are determined for a large number of samples at the
confidence level 0,95 for example, the proportion of those intervals which will contain at least the
desired fraction of the population will be close to 95 %.
9) Table 1 and Table 2 are applicable to the case where the standard deviation for the population is
known (the mean being unknown); Table 3 and Table 4 to the case where the mean and the standard
deviation are unknown.
Where the mean and the standard deviation having respectively the values m and Ö are known, the
distribution of the characteristic under investigation (assumed to be normal) is fully determined; there
is exactly a proportion p of the population:
— on the right side of m – upÖ
one-sided intervals
— on the left side of m + upÖ
— between m – u(1 + p)/2 Ö and m + u(1 + p)/2 Ö: two-sided interval
4)
In preparation.
© BSI 02-2000 1
BS 2846-3:1975
Cx =
2 © BSI 02-2000
BS 2846-3:1975
Cx =
© BSI 02-2000 3
BS 2846-3:1975
n=
Cx =
1–µ=
k2 (n, p, 1 – µ) s = 6)
k2 (n, p, 1 – µ) =
Results
a) One-sided interval “to the left”
There is a probability 1 – ! that at least a proportion p of the population is above:
Ls = x + k2 (n, p, 1 – µ) s =
b) One-sided interval “to the right”
There is a probability 1 – µ that at least a proportion p of the population is above:
Li = x – k2 (n, p, 1 – µ) s =
1) A numerical example is given in section 2 of this International Standard: example No. 3
2) See paragraph 6 of General remarks.
3) See paragraph 7 of General remarks.
4) See paragraph 1 of General remarks.
5) See paragraph 8 of General remarks.
6)
The values of k2 (n, p, 1 – µ) can be read from Table 7 for different values of n, and for
p = 0,90; 0,95; 0,99
1 – µ = 0,95 and 0,99
4 © BSI 02-2000
BS 2846-3:1975
n=
Cx =
1–µ=
k½2 (n, p, 1 – µ) s = 6)
k ′2 (n, p, 1 – µ) =
Results
There is a probability 1 – ! that at least a proportion p of the population is included between the limits7):
Li = x – k ′2 (n, p, 1 – µ) s =
Ls = x + k ′2 (n, p, 1 – µ) s =
1)
A numerical example is given in section 2 of this International Standard: example No. 4
2)
See paragraph 6 of General remarks.
3)
See paragraph 7 of General remarks.
4)
See paragraph 1 of General remarks.
5)
See paragraph 8 of General remarks.
6) The values of k ′ (n, p, 1 – µ) can be read from Table 8 for different values of n, and for
2
p = 0,90; 0,95; 0,99
1 – µ = 0,95 and 0,99
7)
These limits are symmetrical about x but they are not “symmetrical in probability”. It is not true that at the confidence
level 1 – µ, a proportion not exceeding (1 – p)/2 of the population is below Li and a proportion not exceeding (1 – p)/2 is above Ls.
© BSI 02-2000 5
BS 2846-3:1975
Section 2. Examples
Introductory remarks
Table 1 to Table 4 will be illustrated by examples using the numerical values of ISO 2854
(section 2, paragraph 1 of the introductory remarks, Table X, yarn 2): 12 measurements of the breaking
load of cotton yarn. It should be noted that the number of observations n = 12 given here for these examples
is considerably lower than the one recommended in ISO 2062, Textiles — Yarn from packages — Method
for determination of breaking load and elongation at the breaking load of single strands — (CRL, CRE and
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
CRT testers).
The unit of measurement used to express the numerical data and the results of calculations in the different
examples is the centinewton.
x
228,6
232,7
238,8
317,2
315,8
275,1
222,2
236,7
224,7
251,2
210,4
270,7
These measurements come from a batch of 12 000 bobbins, from one production job, packed in 120 boxes
each containing 100 bobbins. 12 boxes have been drawn at random from the batch and a bobbin has been
drawn at random from each of these boxes. Test pieces of 50 cm length have been cut from the yarn on these
bobbins, at about 5 m distance from the free end. The tests themselves have been carried out on the central
parts of these test pieces. Previous information enables it to be allowed that the breaking loads measured
in these conditions have virtually a normal distribution.
These results yield the following:
Sample size:
n = 12
Sum of the observed values:
Cx = 3 024,1
Mean:
Estimated variance:
6 © BSI 02-2000
BS 2846-3:1975
The formal presentation of the calculations will be given only for Table 3 (one-sided interval, unknown
variance).
Numerical examples
Example No. 1 — One-sided statistical tolerance interval (known variance, Table 1)
One assumes that the measurements previously obtained have shown that the dispersion is constant from
one batch to another from the same supplier, although the mean is not constant, and is represented by a
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
© BSI 02-2000 7
BS 2846-3:1975
Technical characteristics of the population under investigation — The batch comprises a delivery of
cotton yarn received on 1969-08-03 from the supplier H and consisting of 12 000 bobbins packed
in 120 boxes each containing 100 bobbins.
Technical characteristics of the sample units — 12 boxes have been drawn at random from the batch and
one bobbin has been taken at random from each of these boxes. Test pieces of 50 cm length have been cut
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
from the yarn on these bobbins, at about 5 m distance from the free end. The tests themselves have been
carried out on the central parts of these test pieces.
Eliminated observations: none
Statistical data Calculations
Sample size:
n = 12
Cx = 3 024,1
8 © BSI 02-2000
BS 2846-3:1975
Other order statistics besides the extreme values may be used, but only the extremes are used here.
A.1.1 One-side spread
Between the sample size n, the confidence level 1 – µ, and the proportion p of the population above xm
(or below xM) one has the relation:
pn = µ
— If one takes values for n and p, this relation enables 1 – µ to be calculated: there is a probability
at least equal to 1 – µ that a proportion of the population at least equal to p is superior to xm
(or inferior to xM).
— If one takes values for n and 1 – µ, this relation enables p to be calculated: there is a
probability at least equal to 1 – µ that a proportion of the population at least equal to p is superior
to xm (or inferior to xM).
— If one takes values for p and 1 – µ, this relation enables one to determine the minimum sample size
n with which it is possible to assert with a confidence level at least equal to 1 – µ that a proportion at
least equal to p of the population will be superior to the smallest value (or inferior to the largest value)
in a sample of that size.
A.1.2 Two-sided spread
Between the sample size n, the proportion p of the population which falls between xm and xM and the
confidence level 1 – µ one has the relation:
npn – 1 – (n – 1)pn = µ
— If one takes values for n and p, this relation enables 1 – µ to be calculated: there is a probability at
least equal to 1 – µ that a proportion of the population at least equal to p falls between xm and xM.
— If one takes values for n and 1 – µ, this relation enables p to be calculated: there is a probability at
least equal to 1 – µ that a proportion of the population at least equal to p falls between xm and xM.
— If one takes values for p and 1 – µ, this relation enables one to determine the minimum sample
size n with which it is possible to assert with a confidence level at least equal to 1 – µ that a proportion
at least equal to p of the population will fall between the smallest and the largest value in a sample of
that size.
A.2 Numerical example
In a fatigue test by rotational stress carried out on a component of an aeronautical engine, a sample
of 15 items has given the following results (measurement of endurance), shown in ascending order of
values:
x
0,200
0,330
0,450
0,490
0,780
0,920
0,950
0,970
1,040
1,710
2,220
2,275
3,650
7,000
8,800
© BSI 02-2000 9
BS 2846-3:1975
A graphical test, for checking, shows that the hypothesis of normality for the population of components
must almost certainly be rejected (see ISO 2854, section 2, paragraph 4 of the introductory remarks).
The methods for Table 3 and Table 4 for determination of a statistical tolerance interval are therefore
inapplicable.
The extreme values from the sample of n = 15 measurements are:
xm = 0,200 xM = 8,800
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy, (c) BSI
10 © BSI 02-2000
BS 2846-3:1975
1 – µ = 0,95 1 – µ = 0,99
n p = 0,90 p = 0,95 p = 0,99 p = 0,90 p = 0,95 p = 0,99
5 2,02 2,38 3,06 2,32 2,69 3,37
6 1,95 2,32 3,00 2,23 2,59 3,28
7 1,90 2,27 2,95 2,16 2,52 3,21
8 1,86 2,23 2,91 2,10 2,47 3,15
9 1,83 2,19 2,87 2,06 2,42 3,10
10 1,80 2,17 2,85 2,02 2,38 3,06
11 1,78 2,14 2,82 1,98 2,35 3,03
12 1,76 2,12 2,80 1,95 2,32 3,00
13 1,74 2,10 2,78 1,93 2,29 2,97
14 1,72 2,08 2,77 1,90 2,27 2,95
15 1,71 2,07 2,75 1,88 2,25 2,93
16 1,69 2,06 2,74 1,86 2,23 2,91
17 1,68 2,04 2,73 1,85 2,21 2,89
18 1,67 2,03 2,71 1,83 2,19 2,87
19 1,66 2,02 2,70 1,82 2,18 2,86
20 1,65 2,01 2,69 1,80 2,17 2,85
22 1,63 2,00 2,68 1,78 2,14 2,82
24 1,62 1,98 2,66 1,76 2,12 2,80
26 1,60 1,97 2,65 1,74 2,10 2,78
28 1,59 1,96 2,64 1,72 2,08 2,77
30 1,58 1,95 2,63 1,71 2,07 2,75
35 1,56 1,92 2,60 1,67 2,04 2,72
40 1,54 1,91 2,59 1,65 2,01 2,69
45 1,53 1,89 2,57 1,63 1,99 2,67
50 1,51 1,88 2,56 1,61 1,97 2,66
60 1,49 1,86 2,54 1,58 1,95 2,63
70 1,48 1,84 2,52 1,56 1,92 2,60
80 1,47 1,83 2,51 1,54 1,91 2,59
90 1,46 1,82 2,50 1,53 1,89 2,57
100 1,45 1,81 2,49 1,51 1,88 2,56
150 1,42 1,78 2,46 1,47 1,83 2,52
200 1,40 1,76 2,44 1,45 1,81 2,49
250 1,39 1,75 2,43 1,43 1,79 2,47
300 1,38 1,74 2,42 1,42 1,78 2,46
400 1,36 1,73 2,41 1,40 1,76 2,44
500 1,36 1,72 2,40 1,39 1,75 2,43
1 000 1,33 1,70 2,38 1,36 1,72 2,40
Z 1,28 1,64 2,33 1,28 1,64 2,33
© BSI 02-2000 11
BS 2846-3:1975
12 © BSI 02-2000
BS 2846-3:1975
1 – µ = 0,95 1 – µ = 0,99
n p = 0,90 p = 0,95 p = 0,99 p = 0,90 p = 0,95 p = 0,99
5 3,41 4,21 5,75
6 3,01 3,71 5,07 4,41 5,41 7,33
7 2,76 3,40 4,64 3,86 4,73 6,41
8 2,58 3,19 4,36 3,50 4,29 5,81
9 2,45 3,03 4,14 3,24 3,97 5,39
10 2,36 2,91 3,98 3,05 3,74 5,08
11 2,28 2,82 3,85 2,90 3,56 4,83
12 2,21 2,74 3,75 2,77 3,41 4,63
13 2,16 2,67 3,66 2,68 3,29 4,47
14 2,11 2,61 3,59 2,59 3,19 4,34
15 2,07 2,57 3,52 2,52 3,10 4,22
16 2,03 2,52 3,46 2,46 3,03 4,12
17 2,00 2,49 3,41 2,41 2,96 4,04
18 1,97 2,45 3,37 2,36 2,91 3,96
19 1,95 2,42 3,33 2,32 2,86 3,89
20 1,93 2,40 3,30 2,28 2,81 3,83
22 1,89 2,35 3,23 2,21 2,73 3,73
24 1,85 2,31 3,18 2,15 2,66 3,64
26 1,82 2,27 3,13 2,10 2,60 3,56
28 1,80 2,24 3,09 2,06 2,55 3,50
30 1,78 2,22 3,06 2,03 2,52 3,45
35 1,73 2,17 2,99 1,96 2,43 3,33
40 1,70 2,13 2,94 1,90 2,37 3,25
45 1,67 2,09 2,90 1,86 2,31 3,18
50 1,65 2,07 2,86 1,82 2,27 3,12
60 1,61 2,02 2,81 1,76 2,20 3,04
70 1,58 1,99 2,77 1,72 2,15 2,98
80 1,56 1,97 2,73 1,69 2,11 2,93
90 1,54 1,94 2,71 1,66 2,08 2,89
100 1,53 1,93 2,68 1,64 2,06 2,85
150 1,48 1,87 2,62 1,57 1,97 2,74
200 1,45 1,84 2,57 1,52 1,92 2,68
250 1,43 1,81 2,54 1,50 1,89 2,64
300 1,42 1,80 2,52 1,48 1,87 2,61
400 1,40 1,78 2,49 1,45 1,84 2,57
500 1,39 1,76 2,48 1,43 1,81 2,54
1 000 1,35 1,73 2,43 1,38 1,76 2,47
Z 1,28 1,64 2,33 1,28 1,64 2,33
© BSI 02-2000 13
BS 2846-3:1975
1 – µ = 0,95 1 – µ = 0,99
n p = 0,90 p = 0,95 p = 0,99 p = 0,90 p = 0,95 p = 0,99
5 4,28 5,08 6,63 6,61 7,86 10,26
6 3,71 4,41 5,78 5,34 6,35 8,30
7 3,37 4,01 5,25 4,61 5,49 7,19
8 3,14 3,73 4,89 4,15 4,94 6,47
9 2,97 3,53 4,63 3,82 4,55 5,97
10 2,84 3,38 4,43 3,58 4,27 5,59
11 2,74 3,26 4,28 3,40 4,05 5,31
12 2,66 3,16 4,15 3,25 3,87 5,08
13 2,59 3,08 4,04 3,13 3,73 4,89
14 2,53 3,01 3,96 3,03 3,61 4,74
15 2,48 2,95 3,88 2,95 3,51 4,61
16 2,44 2,90 3,81 2,87 3,41 4,49
17 2,40 2,86 3,75 2,81 3,35 4,39
18 2,37 2,82 3,70 2,75 3,28 4,31
19 2,34 2,78 3,66 2,70 3,22 4,23
20 2,31 2,75 3,62 2,66 3,17 4,16
22 2,26 2,70 3,54 2,58 3,08 4,04
24 2,23 2,65 3,48 2,52 3,00 3,95
26 2,19 2,61 3,43 2,47 2,94 3,87
28 2,16 2,58 3,39 2,43 2,89 3,79
30 2,14 2,55 3,35 2,39 2,84 3,73
35 2,09 2,49 3,27 2,31 2,75 3,61
40 2,05 2,45 3,21 2,25 2,68 3,52
45 2,02 2,41 3,17 2,20 2,62 3,44
50 2,00 2,38 3,13 2,16 2,58 3,39
60 1,96 2,33 3,07 2,10 2,51 3,29
70 1,93 2,30 3,02 2,06 2,45 3,23
80 1,91 2,27 2,99 2,03 2,41 3,17
90 1,89 2,25 2,96 2,00 2,38 3,13
100 1,87 2,23 2,93 1,98 2,36 3,10
150 1,83 2,18 2,86 1,91 2,27 2,98
200 1,80 2,14 2,82 1,87 2,22 2,92
250 1,78 2,12 2,79 1,84 2,19 2,88
300 1,77 2,11 2,77 1,82 2,17 2,85
400 1,75 2,08 2,74 1,79 2,14 2,81
500 1,74 2,07 2,72 1,78 2,12 2,78
1 000 1,71 2,04 2,68 1,74 2,07 2,72
Z 1,64 1,96 2,58 1,64 1,96 2,58
14 © BSI 02-2000
BS 2846-3:1975
© BSI 02-2000 15
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy
16
BS 2846-3:1975
NOMOGRAPH 1 — One-sided statistical tolerance interval
© BSI 02-2000
Licensed Copy: :FULLNAME, : DATE, Uncontrolled Copy
© BSI 02-2000
BS 2846-3:1975
17
BS 2846-3:
1975
ISO 3207:1975 BSI — British Standards Institution
BSI is the independent national body responsible for preparing
British Standards. It presents the UK view on standards in Europe and at the
international level. It is incorporated by Royal Charter.
Revisions
It is the constant aim of BSI to improve the quality of our products and services.
We would be grateful if anyone finding an inaccuracy or ambiguity while using
this British Standard would inform the Secretary of the technical committee
responsible, the identity of which can be found on the inside front cover.
Tel: 020 8996 9000. Fax: 020 8996 7400.
BSI offers members an individual updating service called PLUS which ensures
that subscribers automatically receive the latest editions of standards.
Buying standards
Orders for all BSI, international and foreign standards publications should be
addressed to Customer Services. Tel: 020 8996 9001. Fax: 020 8996 7001.
Information on standards
Copyright
Copyright subsists in all BSI publications. BSI also holds the copyright, in the
UK, of the publications of the international standardization bodies. Except as
permitted under the Copyright, Designs and Patents Act 1988 no extract may be
reproduced, stored in a retrieval system or transmitted in any form or by any
means – electronic, photocopying, recording or otherwise – without prior written
permission from BSI.
This does not preclude the free use, in the course of implementing the standard,
of necessary details such as symbols, and size, type or grade designations. If these
details are to be used for any other purpose than implementation then the prior
written permission of BSI must be obtained.