Statistical Interpretation of Data - : Guide To

BRITISH STANDARD BS 2846-3:
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
The Advisory Committee on Statistical Methods, under whose supervision this

British Standard was prepared, consists of representatives from the following
Government departments and scientific and industrial organizations.
Biometrika
British National Committee for Surface Active Agents
British Paper and Board Industry Federation (PIF)
Economist Intelligence Unit

Electronic Components Board
Electronic Semiconductor Manufacturers’ Association
European Organization for Quality Control
Institute of Petroleum
Institute of Quality Assurance
Institute of Statisticians
Institution of Production Engineers
Ministry of Defence
National Coal Board
Post Office
Royal Statistical Society
Society of Analytical Chemists
University of Essex
This British Standard, having

been approved by the Advisory
Committee on Statistical
Methods, was published
under the authority of the
Executive Board on Amendments issued since publication
31 October 1975
Amd. No. Date of issue Comments
© BSI 02-2000
2775 March 1979 Indicated by a sideline in the margin

The following BSI references
relate to the work on this
standard:
Committee reference OC/8/2
Draft for approval ISO/DIS 3207
ISBN 0 580 08896 0

BS 2846-3:1975
Contents
Page
Co-operating organizations Inside front cover
Foreword ii
Section 1. Formal presentation of results
General remarks 1
Section 2. Examples
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
results. International agreement has already been reached on the specification of

several techniques and it is proposed to publish these as Parts of a revised
BS 28461), as follows:
Statistical interpretation of data
— Part 1: The routine analysis of quantitative data;
— Part 2: Estimation of the mean — Confidence interval (ISO 2602);
— Part 3: Determination of a statistical tolerance interval (ISO 3207);
— Part 4: Techniques of estimation and tests relating to means and variances
(ISO/DIS 2854)2);
— Part 5: Efficiency of tests relating to means and variances (ISO/DIS 3494)2);
— Part 6: Comparison of two means in the case of paired observations
(ISO/DIS 3301)2).
There is often confusion over the terms “statistical tolerance interval” and
“tolerance zone”. Although closely linked they are not synonymous. The first term
relates to a statistical concept outlined on the next page. The latter term,
however, relates generally to a limiting (physical) requirement specified for, or
imposed on, a particular characteristic (e.g. a dimension of an item may be
specified as 50 ± 2 mm, giving a tolerance zone of 48 mm to 52 mm).
A “statistical tolerance interval”, on the other hand, is a function of inherent
variability the limits of which are often referred to as “natural limits of a process”.
As such it relates directly to “process capability”.
In formal terms a “tolerance zone” is the zone of values in which a measurable
characteristic is in conformity with its specification. A “statistical tolerance
interval” is an interval for which it can be stated with a given level of confidence
that it contains at least a specified proportion of the population (totality of items).
This Part of the standard thus provides techniques by which process capabilities
can be established to enable tolerance zones to be specified objectively (e.g. if a
tolerance zone is specified which lies inside a statistical tolerance interval then
out-of-specification commodities are unavoidable).
Statistical tolerance intervals can be established by either assuming or not
assuming a particular distributional form for the characteristic being studied. In
the latter case the intervals are often called “distribution-free intervals”. The
main advantage of taking into account the distribution of a characteristic is that
the number of determinations (sample size) required to provide an interval
similar to the distribution-free interval is, in general, significantly smaller. The
disadvantage is that the statistical tolerance interval is often critically dependent
on the distribution assumed (i.e. gross errors can result in the size of the interval
and level of confidence if the assumption is false).
1) BS 2846:1957 “The reduction and presentation of experimental results” by J T Richardson, BSc.

2) In course of preparation.
ii © BSI 02-2000
BS 2846-3:1975
Section 1 provides procedures for determining statistical tolerance intervals

based on the assumption that the distribution of the characteristic under study is
“normal”. This subject is discussed briefly in Parts 1 and 4 of this standard and
in more detail in BS . . . .3) “Tests for departure from normality”. Following the
publication of Addendum 1 to ISO 3207 in 1978 which provided graphical
representations of Table 9 and Table 10 in the form of two nomographs,
Amendment No. 1 was issued to this British Standard to incorporate the
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.
© BSI 02-2000 iii

BS 2846-3:1975
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.
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
Section 1. Formal presentation of results
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”.
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
where up is the fractile of order p of the standardized normal variate.

Numerical values of up are in these cases to be read on the bottom line of Table 5 and Table 6.
10) The calculations can often be very much simplified by making a change in origin and/or in unit.
4)
In preparation.
© BSI 02-2000 1
BS 2846-3:1975
Table 1 — One-sided statistical tolerance interval (known variance)1)

Statistical data Calculations

Sample size:
n=
Sum of the observed values:
Cx =
Known value of the variance of the population:

Ö2 =
whence the standard deviation:
Ö= k1 (n, p, 1 – µ) Ö = 6)
Proportion of the population selected for the statistical tolerance
interval4):
p=
Chosen confidence level5):
1–µ=
k1 (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 + k1 (n, p, 1 – µ) Ö =
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 – k1 (n, p, 1 – µ) Ö =
1)
A numerical example is given in section 2 of this International Standard: example No. 1
2)
See paragraph 6 of General remarks.
3)
4)
5)
6) The values of k1 (n, p, 1 – µ) can be read directly from Table 5 for different values of n, and for
p = 0,90; 0,95; 0,99
1 – µ = 0,95 and 0,99
2 © BSI 02-2000
BS 2846-3:1975
Table 2 — Two sided statistical tolerance interval (known variance)1)


Sample size:
n=
Cx =
Known value of the variance of the population:

B2 =
whence the standard deviation:
B= k½1 (n, p, 1 – µ) Ö = 6)
Proportion of the population selected for the

statistical tolerance interval4):
p=
Chosen confidence level5):
1–µ=
k½1 (n, p, 1 – µ) =
Results
There is a probability 1 – µ that at least a proportion p of the population is included between the limits7):
Li = x – k½1 (n, p, 1 – µ) Ö =
Ls = x + k½1 (n, p, 1 – µ) Ö =
1)
A numerical example is given in section 2 of this International Standard: example No. 2.
2)
3)
4)
see paragraph 1 of General remarks.
5)
see paragraph 8 of General remarks.
6) The values of k½ (n, p, 1 – µ) can be read from Table 6 for different values of n, and for
1
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 3
BS 2846-3:1975
Table 3 — One-sided statistical tolerance interval (unknown variance)1)


Sample size:
n=
Cx =
Sum of the squares of the observed values:

Cx2 =
p= (estimation of the standard deviation Ö)

Chosen confidence level : 5)
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.
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
Table 4 — Two-sided statistical tolerance interval (unknown variance)1)


Sample size:
n=
Cx =

Cx2 =
p= (estimation of the standard deviation Ö)

Chosen confidence level : 5)
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)
3)
4)
5)
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
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
Cx = 3 024,1
Mean:

Cx2 = 775 996,09
Sum of the squares of the differences from the mean value:
Estimated variance:
Estimated standard deviation:

s = 35,5
It is also known from experience that in the same batch the breaking loads are distributed according to a
pattern very close to normal distribution.
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
standard deviation Ö = 33,15.

One wishes to calculate the limit Li such that it is possible to assert with a confidence level 1 – µ = 0,95 that
in a proportion at least equal to 0,95 (95 %) the breaking load of the items likely to be taken as samples
from the batch and measured in the same conditions is above Li.
Table 5 gives:
k1 (12; 0,95; 0,95) = 2,12
whence:
Li = x – k1Ö = 252,0 – 2,12 × 33,15 = 181,7
One would, of course, obtain a smaller limit Li if one took a higher proportion of the population
(for example 99 %) and/or a higher confidence level (for example 1 – µ = 0,99).
Example No. 2 — Two-sided statistical tolerance interval (known variance, Table 2)
In the same conditions as in example No. 1, one wishes to calculate the limits Li and Ls such that it is
possible to assert with a confidence level 1 – µ = 0,95 that in a proportion of the batch at least equal to
p = 0,90 (90 %) the breaking load falls between Li and Ls.
Table 6 gives:
k ′1 (12; 0,90; 0,95) = 1,89
whence:
Li = x – k ′1 Ö = 252,0 – 1,89 × 33,15 = 189,3
Ls = x + k ′1 Ö = 252,0 + 1,89 × 33,15 = 314,7
An attempt must be made to eliminate the suspicion that 5 % (at the maximum) of the population lies to
the left of Li and similarly, to the right of Ls.
It has been seen, in connection with example No. 1, that the limit to the left of which not more than a
maximum of 5 % of the population lies (minimum of 95 % to the right of Li) is Li = 181,7.
Example No. 3 — One-sided statistical tolerance interval (unknown variance, Table 3)
Here, one assumes that the standard deviation of the population is unknown and has to be estimated from
the sample. One will take the same conditions as for the case where the standard deviation is known
(example No. 1) — thus, p = 0,95 and 1 – µ = 0,95.
© BSI 02-2000 7
BS 2846-3:1975
The presentation of the results is given in detail below.
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
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
Sample size:
n = 12
Cx = 3 024,1
Cx2 = 775 996,09 = 1 263,4

Proportion of the population selected for the statistical Ö* = s = 1 263,4 = 35,5
tolerance interval:
p = 0,95 (95 %) k2 (12; 0,95; 0,95) s = 97,3
Chosen confidence level:
1 – µ = 0,95
k2 (12; 0,95; 0,95) = 2,74
(value read from Table 7)
Results
It is possible to assert with a confidence level of 0,95 that a proportion at least equal to 0,95 (95 %) of the
breaking loads for the batch is above:
Li = 252,0 – 2,74 × 35,5 = 154,7
It will be noted that the value of Li is smaller than in example No. 1 (known variance), because the
use of s, random estimation of Ö, produces a higher value of the coefficient k (2,74 instead of 2,12).
Example No. 4 — Two-sided statistical tolerance interval (unknown variance, Table 4)
In the same conditions as in example No. 3, one wishes to calculate the limits Li and Ls such that it is
possible to assert with a confidence level 1 – µ = 0,95 that in a proportion of the batch at least equal to
p = 0,90 (90 %) the breaking load falls between Li and Ls.
Table 8 gives:
k½2 (12; 0,90; 0,95) = 2,66
whence:
Li = x – k′2 s = 252,0 – 2,66 × 35,5 = 157,6
Ls = x – k′2 s = 252,0 + 2,66 × 35,5 = 346,4
It will be noted that the value of Li is smaller and the value of Ls higher than in example No. 2
(known variance) because the use of s, random estimation of Ö, produces a higher value of the
coefficient k (2,66 instead of 1,89).
We have to pay the penalty for not knowing the population standard deviation Ö and the extending of the
interval takes this into account, but of course if we are not quite sure that the value Ö = 33,15 used in
examples Nos. 1 and 2 is correct, it may be wiser to use the estimate, s, in Table 3 or Table 4.
8 © BSI 02-2000
BS 2846-3:1975
Annex A Case of any distribution

The method described here uses extreme values xm and xM in the sample.
A.1 Introductory remarks
In the case of a continuous distribution represented by a random and simple sample of any size n and where
all hypothesis is missing, information about the degree of spread of the distribution can be obtained from
the extreme values of the sample (xm the smallest, xM the largest) or from only one of these.
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
One takes a confidence level 1 – µ = 0,95.

a1) What is the maximum proportion of the population of components that will fall below xm = 0,200?
Table 9, for 1 – µ = 0,95, gives for the minimum proportion above xm a value p slightly higher
than 0,75 (75 %) — hence, for the maximum proportion below xm, a value 1 – p slightly lower
than 0,25 (25 %).
a2) What sample size must be taken to make it possible to assert (at a confidence level 0,95) that a
proportion at least p = 0,90 (90 %) of the population of components will be found below the largest of
the values from that sample?
Table 9, for 1 – µ = 0,95 and p = 0,90, gives n = 29.
b1) At the confidence level 0,95, what is the minimum proportion of the population of components that
falls between xm = 0,200 and xM = 8,800?
Table 10, for 1 – µ = 0,95 and n = 15, gives p slightly below 0,75 (75 %).
b2) What sample size must be taken to make it possible to assert (at a confidence level 0,95) that a
proportion of at least p = 0,90 (90 %) of the population of components will be found to fall between the
smallest and the largest of the values from that sample?
Table 10, for 1 – µ = 0,95 and p = 0,90 gives n = 46.
10 © BSI 02-2000
BS 2846-3:1975
Annex B Statistical tables

Table 5 — One-sided statistical tolerance interval, Ö known,
m unknown
x + k 1 Ö or x – k 1 Ö
Values of the coefficient k1 (n, p, 1 – µ)
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
Table 6 — Two sided statistical tolerance interval, Ö known, m unknown

x ± k′ Ö1
Values of the coefficient k ′1 (n, p, 1 – µ)

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 2,16 2,52 3,20 2,44 2,80 3,48

6 2,09 2,45 3,13 2,34 2,70 3,38
7 2,04 2,40 3,07 2,26 2,62 3,30
8 2,00 2,35 3,02 2,20 2,56 3,24
9 1,96 2,31 2,98 2,15 2,51 3,19
10 1,93 2,28 2,95 2,11 2,47 3,14
11 1,91 2,26 2,93 2,07 2,43 3,10
12 1,89 2,24 2,90 2,04 2,40 3,07
13 1,87 2,22 2,88 2,02 2,37 3,04
14 1,86 2,20 2,86 1,99 2,35 3,02
15 1,84 2,19 2,85 1,97 2,33 3,00
16 1,83 2,17 2,83 1,95 2,31 2,98
17 1,82 2,16 2,82 1,94 2,29 2,96
18 1,81 2,15 2,81 1,92 2,27 2,94
19 1,80 2,14 2,80 1,91 2,26 2,93
20 1,80 2,13 2,79 1,90 2,25 2,91
22 1,78 2,12 2,77 1,88 2,22 2,89
24 1,77 2,11 2,76 1,86 2,20 2,87
26 1,76 2,10 2,74 1,84 2,19 2,85
28 1,75 2,09 2,73 1,83 2,17 2,83
30 1,75 2,08 2,72 1,82 2,16 2,82
35 1,73 2,06 2,71 1,79 2,13 2,79
40 1,72 2,05 2,69 1,78 2,11 2,76
45 1,71 2,04 2,68 1,76 2,10 2,74
50 1,71 2,03 2,67 1,75 2,08 2,73
60 1,70 2,02 2,65 1,73 2,06 2,71
70 1,69 2,01 2,64 1,72 2,05 2,69
80 1,68 2,01 2,63 1,71 2,04 2,68
90 1,68 2,00 2,63 1,70 2,03 2,67
100 1,68 2,00 2,62 1,70 2,02 2,66
150 1,67 1,98 2,61 1,68 2,00 2,63
200 1,66 1,98 2,60 1,67 1,99 2,62
250 1,66 1,97 2,60 1,67 1,99 2,61
300 1,66 1,97 2,59 1,66 1,98 2,60
400 1,65 1,97 2,59 1,66 1,98 2,60
500 1,65 1,97 2,59 1,66 1,97 2,59
1 000 1,65 1,96 2,58 1,65 1,96 2,58
Z 1,64 1,96 2,58 1,64 1,96 2,58
12 © BSI 02-2000
BS 2846-3:1975
Table 7 — One-sided statistical tolerance interval, m, Ö unknown
Values of the coefficient k2 (n, p, 1 – µ)

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
Table 8 — Two sided statistical tolerance interval, m, Ö unknown
Values of the coefficient k½2 (n, p, 1 – µ)

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
Table 9 — Non-parametric one-sided statistical tolerance

intervals — Sample size n for a proportion p at confidence level 1 – µ
pn = µ
p
1–µ
0,50 0,75 0,90 0,95 0,99 0,999
0,50 1 3 7 14 60 693
0,75 3 5 14 28 138 1 386

0,90 4 9 22 45 230 2 302
0,95 5 11 29 59 299 2 995
0,99 7 17 44 90 459 4 603
0,999 10 25 66 135 688 6 905
Table 10 — Non-parametric two-sided tolerance intervals — Sample size

n for a proportion p at confidence level 1 – µ
npn – 1 – (n – 1) pn = µ
p
1–µ
0,50 0,75 0,90 0,95 0,99 0,999
0,50 3 7 17 34 168 1 679
0,75 5 10 27 53 269 2 692
0,90 7 15 38 77 388 3 889
0,95 8 18 46 93 473 4 742
0,99 11 24 64 130 662 6 636
0,999 14 33 89 181 920 9 230
© 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
NOMOGRAPH 2 — Two-sided statistical tolerance interval
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
British Standards are updated by amendment or revision. Users of

British Standards should make sure that they possess the latest amendments or
editions.
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.
In response to orders for international standards, it is BSI policy to supply the

BSI implementation of those that have been published as British Standards,
unless otherwise requested.
Information on standards
BSI provides a wide range of information on national, European and

international standards through its Library and its Technical Help to Exporters
Service. Various BSI electronic information services are also available which give
details on all its products and services. Contact the Information Centre.
Tel: 020 8996 7111. Fax: 020 8996 7048.
Subscribing members of BSI are kept up to date with standards developments

and receive substantial discounts on the purchase price of standards. For details
of these and other benefits contact Membership Administration.
Tel: 020 8996 7002. Fax: 020 8996 7001.
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.
If permission is granted, the terms may include royalty payments or a licensing

agreement. Details and advice can be obtained from the Copyright Manager.
BSI Tel: 020 8996 7070.
389 Chiswick High Road
London
W4 4AL

Statistical Interpretation of Data - : Guide To

Uploaded by

Copyright:

Available Formats

You might also like

Statistical Interpretation of Data - : Guide To

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Statistical Interpretation of Data - : Guide To

Uploaded by

Copyright:

Available Formats

BRITISH STANDARD BS 2846-3:

The Advisory Committee on Statistical Methods, under whose supervision this

Economist Intelligence Unit

This British Standard, having

2775 March 1979 Indicated by a sideline in the margin

ISBN 0 580 08896 0

results. International agreement has already been reached on the specification of

1) BS 2846:1957 “The reduction and presentation of experimental results” by J T Richardson, BSc.

Section 1 provides procedures for determining statistical tolerance intervals

© BSI 02-2000 iii

Section 1. Formal presentation of results

where up is the fractile of order p of the standardized normal variate.

Table 1 — One-sided statistical tolerance interval (known variance)1)

Statistical data Calculations

Known value of the variance of the population:

Table 2 — Two sided statistical tolerance interval (known variance)1)

Statistical data Calculations

Known value of the variance of the population:

Proportion of the population selected for the

Table 3 — One-sided statistical tolerance interval (unknown variance)1)

Statistical data Calculations

Sum of the observed values:

Sum of the squares of the observed values:

p= (estimation of the standard deviation Ö)

Table 4 — Two-sided statistical tolerance interval (unknown variance)1)

Statistical data Calculations

Sum of the observed values:

Sum of the squares of the observed values:

p= (estimation of the standard deviation Ö)

Sum of the squares of the observed values:

Estimated standard deviation:

standard deviation Ö = 33,15.

The presentation of the results is given in detail below.

Sum of the observed values:

Sum of the squares of the observed values:

Cx2 = 775 996,09 = 1 263,4

Annex A Case of any distribution

One takes a confidence level 1 – µ = 0,95.

Annex B Statistical tables

Table 6 — Two sided statistical tolerance interval, Ö known, m unknown

Values of the coefficient k ′1 (n, p, 1 – µ)

5 2,16 2,52 3,20 2,44 2,80 3,48

Table 7 — One-sided statistical tolerance interval, m, Ö unknown

Values of the coefficient k2 (n, p, 1 – µ)

Table 8 — Two sided statistical tolerance interval, m, Ö unknown

Values of the coefficient k½2 (n, p, 1 – µ)

Table 9 — Non-parametric one-sided statistical tolerance

0,75 3 5 14 28 138 1 386

Table 10 — Non-parametric two-sided tolerance intervals — Sample size

NOMOGRAPH 2 — Two-sided statistical tolerance interval

British Standards are updated by amendment or revision. Users of

In response to orders for international standards, it is BSI policy to supply the

BSI provides a wide range of information on national, European and

Subscribing members of BSI are kept up to date with standards developments

If permission is granted, the terms may include royalty payments or a licensing

You might also like