This international standard was developed in accordance with internationally recognized principles on standardization established in the Decision on Principles

for the
Development of International Standards, Guides and Recommendations issued by the World Trade Organization Technical Barriers to Trade (TBT) Committee.

Designation: E178 − 21 An American National Standard

Standard Practice for

Dealing With Outlying Observations1
This standard is issued under the fixed designation E178; the number immediately following the designation indicates the year of
original adoption or, in the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. A
superscript epsilon (´) indicates an editorial change since the last revision or reapproval.

1. Scope 3.1.2.1 Discussion—In this practice, xk is used to denote

1.1 This practice covers outlying observations in samples order statistics in place of x(k), to simplify the notation.
and how to test the statistical significance of outliers. 3.1.3 outlier—see outlying observation.
1.2 The system of units for this standard is not specified. 3.1.4 outlying observation, n—an extreme observation in
Dimensional quantities in the standard are presented only as either direction that appears to deviate markedly in value from
illustrations of calculation methods. The examples are not other members of the sample in which it appears.
binding on products or test methods treated. 3.1.4.1 Discussion—The identification of a value as
1.3 This standard does not purport to address all of the outlying, and therefore a doubtful observation, is a judgement
safety concerns, if any, associated with its use. It is the of the analyst and can be made before any statistical test.
responsibility of the user of this standard to establish appro-
priate safety, health, and environmental practices and deter- 4. Significance and Use
mine the applicability of regulatory limitations prior to use. 4.1 An outlying observation, or “outlier,” is an extreme one
1.4 This international standard was developed in accor- in either direction that appears to deviate markedly from other
dance with internationally recognized principles on standard- members of the sample in which it occurs.
ization established in the Decision on Principles for the 4.2 Statistical rules test the null hypothesis of no outliers
Development of International Standards, Guides and Recom- against the alternative of one or more actual outliers. The
mendations issued by the World Trade Organization Technical procedures covered were developed primarily to apply to the
Barriers to Trade (TBT) Committee. simplest kind of experimental data, that is, replicate measure-
2. Referenced Documents ments of some property of a given material or observations in
a supposedly random sample.
2.1 ASTM Standards:2
E456 Terminology Relating to Quality and Statistics 4.3 A statistical test may be used to support a judgment that
E2586 Practice for Calculating and Using Basic Statistics a physical reason does actually exist for an outlier, or the
statistical criterion may be used routinely as a basis to initiate
3. Terminology action to find a physical cause.
3.1 Definitions—Unless otherwise noted in this standard, all 5. Procedure
terms relating to quality and statistics are defined in Terminol-
5.1 In dealing with an outlier, the following alternatives
ogy E456.
should be considered:
3.1.1 null hypothesis, H0, n—a statement about a parameter
5.1.1 An outlying observation might be the result of gross
of a probability distribution or about the type of probability
deviation from prescribed experimental procedure or an error
distribution, tentatively regarded as true until rejected using a
in calculating or recording the numerical value. When the
statistical hypothesis test. E2586
experimenter is clearly aware that a deviation from prescribed
3.1.2 order statistic x(k), n—value of the kth observed value experimental procedure has taken place, the resultant observa-
in a sample after sorting by order of magnitude. E2586 tion should be discarded, whether or not it agrees with the rest
of the data and without recourse to statistical tests for outliers.
1
This practice is under the jurisdiction of ASTM Committee E11 on Quality and If a reliable correction procedure is available, the observation
Statistics and is the direct responsibility of Subcommittee E11.10 on Sampling / may sometimes be corrected and retained.
Statistics.
Current edition approved June 1, 2021. Published June 2021. Originally 5.1.2 An outlying observation might be merely an extreme
approved in 1961. Last previous edition approved in 2016 as E178 – 16a. DOI: manifestation of the random variability inherent in the data. If
10.1520/E0178-21.
2
this is true, the value should be retained and processed in the
For referenced ASTM standards, visit the ASTM website, www.astm.org, or
contact ASTM Customer Service at [email protected]. For Annual Book of ASTM
same manner as the other observations in the sample. Trans-
Standards volume information, refer to the standard’s Document Summary page on formation of data or using methods of data analysis designed
the ASTM website. for a non-normal distribution might be appropriate.

Copyright © ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. United States

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

1
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
5.1.3 Test units that give outlying observations might be of 6.2 Almost all criteria for outliers are based on an assumed
special interest. If this is true, once identified they should be underlying normal (Gaussian) population or distribution. The
segregated for more detailed study. Outliers may contain null hypothesis that we are testing in every case is that all
important information for a possible root cause analysis and observations in the sample come from the same normal
action on the process or procedure. population. In choosing an appropriate alternative hypothesis
5.2 In many cases, evidence for deviation from prescribed (one or more outliers, separated or bunched, on same side or
procedure will consist primarily of the discordant value itself. different sides, and so forth) it is useful to plot the data as
In such cases it is advisable to adopt a cautious attitude. Use of shown in the dot diagrams of the figures. When the data are not
one of the criteria discussed below will sometimes permit a normally or approximately normally distributed, the probabili-
clearcut decision to be made. ties associated with these tests will be different. The experi-
5.2.1 When the experimenter cannot identify abnormal menter is cautioned against interpreting the probabilities too
conditions, they should report the discordant values and literally.
indicate to what extent they have been used in the analysis of 6.3 Although our primary interest here is that of detecting
the data. outlying observations, some of the statistical criteria presented
5.3 Thus, as part of the over-all process of experimentation, may also be used to test the hypothesis of normality or that the
the process of screening samples for outlying observations and random sample taken come from a normal or Gaussian
acting on them is the following: population. The end result is for all practical purposes the
5.3.1 Physical Reason Known or Discovered for Outlier(s): same, that is, we really wish to know whether we ought to
5.3.1.1 Reject observation(s) and possibly take additional proceed as if we have in hand a sample of homogeneous
observation(s). normal observations.
5.3.1.2 Correct observation(s) on physical grounds. 6.4 One should distinguish between data to be used to
5.3.2 Physical Reason Unknown—Use Statistical Test: estimate a central value from data to be used to assess
5.3.2.1 Reject observation(s) and possibly take additional variability. When the purpose is to estimate a standard
observation(s). deviation, it might be seriously underestimated by dropping too
5.3.2.2 Transform observation(s) to improve fit to a normal many “outlying” observations.
distribution.
5.3.2.3 Use estimation appropriate for non-normal distribu- 7. Recommended Criteria for Single Samples
tions. 7.1 Criterion for a Single Outlier—Sort the n observations
5.3.2.4 Segregate samples for further study. in order of increasing magnitude by x1 ≤ x2 ≤ x3 ≤ ... ≤ xn,
called order statistics. Let the largest value, xn, be the doubtful
6. Basis of Statistical Criteria for Outliers value, that is the largest value. The test criterion, Tn, for a
6.1 In testing outliers, the doubtful observation is included single outlier is as follows:
in the calculation of the numerical value of a sample criterion T n 5 ~ x n 2 x̄ ! /s (1)
(or statistic), which is then compared with a critical value
where:
based on the theory of random sampling to determine whether
the doubtful observation is to be retained or rejected. The x̄ = arithmetic average of all n values, and
critical value is that value of the sample criterion which would s = estimate of the population standard deviation based on
be exceeded by chance with some specified (small) probability the sample data, calculated as follows:
on the assumption that all the observations did indeed consti- 1 n
tute a random sample from a common system of causes, a x̄ 5 ( x
n i51 i
(2)
single parent population, distribution or universe. The specified n n

! !
small probability is called the “significance level” or “percent-
age point” and can be thought of as the risk of erroneously
(
i51
~ x i 2 x̄ ! 2 (x
i51
i
2
2 n·x̄ 2

s5 5
rejecting a good observation. If a real shift or change in the n21 n21
value of an observation arises from nonrandom causes (human
error, loss of calibration of instrument, change of measuring
5 !(
n

i51
xi 2 2 S( Dn

i51
xi
2

/n
(3)
instrument, or even change of time of measurements, and so n21
forth), then the observed value of the sample criterion used will
exceed the “critical value” based on random-sampling theory. If x1 rather than xn is the doubtful value, the criterion is as
Tables of critical values are usually given for several different follows:
significance levels. In particular for this practice, significance T 1 5 ~ x̄ 2 x 1 ! /s (4)
levels 10, 5, and 1 % are used.
The critical values for either case, for the 1, 5, and 10 %
NOTE 1—In this practice, we will usually illustrate the use of the 5 % levels of significance, are given in Table 1.
significance level. Proper choice of level in probability depends on the 7.1.1 The test criterion Tn can be equated to the Student’s t
particular problem and just what may be involved, along with the risk that
one is willing to take in rejecting a good observation, that is, if the test statistic for equality of means between a population with
null-hypothesis stating “all observations in the sample come from the one observation xn and another with the remaining observa-
same normal population” may be assumed correct. tions x1, ... , xn – 1, and the critical value of Tn for significance

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

2
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
TABLE 1 Critical Values for T (One-Sided Test) When Standard
Deviation is Calculated from the Same SampleA
Number of Upper 10 % Upper 5 % Upper 1 %
Observations, Significance Significance Significance
n Level Level Level FIG. 1 Ten Observations of Breaking Strength from Example 1
3 1.1484 1.1531 1.1546
4 1.4250 1.4625 1.4925
5 1.602 1.672 1.749 T 10 5 ~ 596 2 575.2! /8.70 5 2.39
6 1.729 1.822 1.944
7 1.828 1.938 2.097 From Table 1, for n = 10, note that a T10 as large as 2.39
8 1.909 2.032 2.221
9 1.977 2.110 2.323
would occur by chance with probability less than 0.05. In fact,
10 2.036 2.176 2.410 so large a value would occur by chance not much more often
11 2.088 2.234 2.485 than 1 % of the time. Thus, the weight of the evidence is
12 2.134 2.285 2.550
13 2.175 2.331 2.607
against the doubtful value having come from the same popu-
14 2.213 2.371 2.659 lation as the others (assuming the population is normally
15 2.247 2.409 2.705 distributed). Investigation of the doubtful value is therefore
16 2.279 2.443 2.747
17 2.309 2.475 2.785
indicated.
18 2.335 2.504 2.821 7.2 Dixon Criteria for a Single Outlier—An alternative
19 2.361 2.532 2.854
20 2.385 2.557 2.884 system, the Dixon criteria (2),3 based entirely on ratios of
21 2.408 2.580 2.912 differences between the observations may be used in cases
22 2.429 2.603 2.939 where it is desirable to avoid calculation of s or where quick
23 2.448 2.624 2.963
24 2.467 2.644 2.987 judgment is called for. For the Dixon test, the sample criterion
25 2.486 2.663 3.009 or statistic changes with sample size. Table 2 gives the
26 2.502 2.681 3.029 appropriate statistic to calculate and also gives the critical
27 2.519 2.698 3.049
28 2.534 2.714 3.068 values of the statistic for the 1, 5, and 10 % levels of
29 2.549 2.730 3.085 significance. In most situations, the Dixon criteria is less
30 2.563 2.745 3.103 powerful at detecting an outlier than the criterion given in 7.1.
35 2.628 2.811 3.178
40 2.682 2.866 3.240 7.2.1 Example 2—As an illustration of the use of Dixon’s
45 2.727 2.914 3.292 test, consider again the observations on breaking strength given
50 2.768 2.956 3.336 in Example 1. Table 2 indicates use of:
A
Values of T are taken from Grubbs (1),3 Table 1. All values have been adjusted
for division by n – 1 instead of n in calculating s. Use Ref. (1) for higher sample r 11 5 ~ x n 2 x n21 ! / ~ x n 2 x 2! (6)
sizes up to n = 147.
Thus, for n = 10:
r 11 5 ~ x 10 2 x 9 ! / ~ x 10 2 x 2 ! (7)

level α can be approximated using the α/n percentage point of For the measurements of breaking strength above:
Student’s t with n – 2 degrees of freedom. The approximation r 11 5 ~ 596 2 584! / ~ 596 2 570! 5 0.462
is exact for small enough values of α, depending on n, and
Which is a little less than 0.478, the 5 % critical value for n
otherwise a slight overestimate unless both α and n are large:
= 10. Under the Dixon criterion, we should therefore not
t α⁄n,n22 consider this observation as an outlier at the 5 % level of
T n~ α ! # (5)
Π11
2
nt α⁄n,n22
~ 2 1!2
n
21 significance. These results illustrate how borderline cases may
be accepted under one test but rejected under another.
7.1.2 To test outliers on the high side, use the statistic Tn = 7.3 Recursive Testing for Multiple Outliers in Univariate
(xn – x̄ )/s and take as critical value the 0.05 point of Table 1. Samples—For testing multiple outliers in a sample, recursive
To test outliers on the low side, use the statistic T1 = (x̄ – x1)/s application of a test for a single outlier may be used. In
and again take as a critical value the 0.05 point of Table 1. If recursive testing, a test for an outlier, x1 or xn, is first
we are interested in outliers occurring on either side, use the conducted. If this is found to be significant, then the test is
statistic Tn = (xn – x̄ )/s or the statistic T1 = (x̄ – x1)/s whichever repeated, omitting the outlier found, to test the point on the
is larger. If in this instance we use the 0.05 point of Table 1 as opposite side of the sample, or an additional point on the same
our critical value, the true significance level would be twice side. The performance of most tests for single outliers is
0.05 or 0.10. Similar considerations apply to the other tests affected by masking, where the probability of detecting an
given below. outlier using a test for a single outlier is reduced when there are
7.1.3 Example 1—As an illustration of the use of Tn and two or more outliers. Therefore, the recommended procedure is
Table 1, consider the following ten observations on breaking to use a criterion designed to test for multiple outliers, using
strength (in pounds) of 0.104-in. hard-drawn copper wire: 568, recursive testing to investigate after the initial criterion is
570, 570, 570, 572, 572, 572, 578, 584, 596. See Fig. 1. The significant.
doubtful observation is the high value, x10 = 596. Is the value
of 596 significantly high? The mean is x̄ = 575.2 and the 3
The boldface numbers in parentheses refer to a list of references at the end of
estimated standard deviation is s = 8.70. We compute: this standard.

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

3
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
TABLE 2 Dixon Criteria for Testing of Extreme Observation (Single Sample)A
Significance Level (One-Sided Test)
n Criterion
10 % 5% 1%
3 r10 = (x2 − x1)/(xn − x1) if smallest value is suspected; 0.886 0.941 0.988
4 = (xn − xn−1)/(xn − x1) if largest value is suspected 0.679 0.766 0.889
5 0.558 0.642 0.781
6 0.484 0.562 0.698
7 0.434 0.507 0.637
8 r11 = (x2 − x1)/(xn−1 − x1) if smallest value is suspected; 0.480 0.554 0.681
9 = (xn − xn−1)/(xn − x2) if largest value is suspected. 0.440 0.511 0.634
10 0.410 0.478 0.597
11 r21 = (x3 − x1)/(xn−1 − x1) if smallest value is suspected; 0.517 0.575 0.674
12 = (xn − xn−2)/(xn − x2) if largest value is suspected. 0.490 0.546 0.643
13 0.467 0.521 0.617
14 r22 = (x3 − x1)/(xn−2 − x1) if smallest value is suspected; 0.491 0.546 0.641
15 = (xn − xn−2)/(xn − x3) if largest value is suspected. 0.470 0.524 0.618
16 0.453 0.505 0.598
17 0.437 0.489 0.580
18 0.424 0.475 0.564
19 0.412 0.462 0.550
20 0.401 0.450 0.538
21 0.391 0.440 0.526
22 0.382 0.430 0.516
23 0.374 0.421 0.506
24 0.366 0.413 0.497
25 0.359 0.406 0.489
26 0.353 0.399 0.482
27 0.347 0.393 0.474
28 0.342 0.387 0.468
29 0.336 0.381 0.462
30 0.332 0.376 0.456
35 0.311 0.354 0.431
40 0.295 0.337 0.412
45 0.283 0.323 0.397
50 0.272 0.312 0.384
A
x1 # x2 # ... # xn. Original Table in Dixon (2), Appendix. Critical values updated by calculations by Bohrer (3) and Verma-Ruiz (4).

7.4.2 The deviations –1.40 and 1.01 appear to be outliers.

Here the suspected observations lie at each end of the sample.
The mean of the deviations is x̄ = 0.018, the standard deviation
is s = 0.551, and:
w/s 5 @ 1.01 2 ~ 21.40! # /0.551 5 2.41/0.551 5 4.374
From Table 3 for n = 15, we see that the value of w/s = 4.374
FIG. 2 Fifteen Residuals from the Semidiameters of Venus from
Example 3 falls between the critical values for the 1 and 5 % levels, so if
the test were being run at the 5 % level of significance, we
would conclude that this sample contains one or more outliers.
7.4.3 The lowest measurement, –1.40, is 1.418 below the
sample mean, and the highest measurement, 1.01, is 0.992
7.4 Criterion for Two Outliers on Opposite Sides of a
above the mean. Since these extremes are not symmetric about
Sample—In testing the least and the greatest observations the mean, either both extremes are outliers, or else only –1.40
simultaneously as probable outliers in a sample, use the ratio of is an outlier. That –1.40 is an outlier can be verified by use of
sample range to sample standard deviation test of David, the T1 statistic. We have:
Hartley, and Pearson (5):
T 1 5 ~ x̄ 2 x 1 ! /s 5 @ 0.018 2 ~ 21.40! # /0.551 5 2.574
w/s 5 ~ x n 2 x 1 ! /s (8)
This value is greater than the critical value for the 5 % level,
The significance levels for this sample criterion are given in 2.409 from Table 1, so we reject –1.40. Since we have decided
Table 3. Alternatively, the largest residuals test of Tietjen and that –1.40 should be rejected, we use the remaining 14
Moore (7.5) could be used. observations and test the upper extreme 1.01, either with the
7.4.1 Example 3—This classic set consists of a sample of 15 criterion:
observations of the vertical semidiameters of Venus made by T n 5 ~ x n 2 x̄ ! /s (9)
Lieutenant Herndon in 1846 (6). In the reduction of the
observations, Prof. Pierce found the following residuals (in or with Dixon’s r22. Omitting –1.40 and renumbering the
seconds of arc) which have been arranged in ascending order of observations, we compute:
magnitude. See Fig. 2, above. x̄ 5 1.67/14 5 0.119, s 5 0.401

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

4
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
TABLE 3 Critical ValuesA (One-Sided Test) for w/s (Ratio of Now relabel the original observations x1, x2, ..., xn as z’s in
Range to Sample Standard Deviation)
such a manner that zi is that x whose ri is the ith smallest
Number of 10 % 5% 1%
Observations, Significance Significance Significance
absolute residual above. This now means that z1 is that
n Level Level Level observation x which is closest to the mean and that zn is the
3 1.9973 1.9993 2.0000 observation x which is farthest from the mean. The Tietjen-
4 2.409 2.429 2.445 Moore statistic for testing the significance of the k largest
5 2.712 2.755 2.803
6 2.949 3.012 3.095 residuals is then:
7
8
9
3.143
3.308
3.449
3.222
3.399
3.552
3.338
3.543
3.720
Ek 5 F( ~
n2k

i51
z i 2 z̄ k ! 2 /
n

( ~ z 2 z̄ !
i51
i
2
G (11)
10 3.574 3.685 3.875
11 3.684 3.803 4.011 where:
12 3.782 3.909 4.133
n2k
13 3.871 4.005 4.244
14 3.952 4.092 4.344 z̄ k 5 ( z /~n 2 k!
i51
i (12)
15 4.025 4.171 4.435
16 4.093 4.244 4.519 is the mean of the (n − k) least extreme observations and z̄ is
17 4.156 4.311 4.597
the mean of the full sample. Percentage points of Ek in Table 4
18 4.214 4.374 4.669
19 4.269 4.433 4.736 were computed by simulation.
20 4.320 4.487 4.799
21 4.368 4.539 4.858 7.5.1 Example 4—Applying this test to the Venus semidi-
22 4.413 4.587 4.913 ameter residuals data in Example 3, we find that the total sum
23 4.456 4.633 4.965
24 4.497 4.676 5.015
of squares of deviations for the entire sample is 4.24964.
25 4.535 4.717 5.061 Omitting –1.40 and 1.01, the suspected two outliers, we find
26 4.572 4.756 5.106 that the sum of squares of deviations for the reduced sample of
27 4.607 4.793 5.148
28 4.641 4.829 5.188
13 observations is 1.24089. Then E2 = 1.24089/4.24964 =
29 4.673 4.863 5.226 0.292, and by using Table 4, we find that this observed E2 is
30 4.704 4.895 5.263 slightly smaller than the 5 % critical value of 0.317, so that the
35 4.841 5.040 5.426
40 4.957 5.162 5.561
E2 test would reject both of the observations, –1.40 and 1.01.
45 5.057 5.265 5.674
50 5.144 5.356 5.773
7.6 Criterion for Two Outliers on the Same Side of the
A
Sample—Where the two largest or the two smallest observa-
Each entry calculated by 50 000 000 simulations.
tions are probable outliers, employ a test provided by Grubbs
(8, 9) which is based on the ratio of the sample sum of squares
when the two doubtful values are omitted to the sample sum of
squares when the two doubtful values are included. In illus-
and: trating the test procedure, we give the following Examples 5
and 6.
T 14 5 ~ 1.01 2 0.119! /0.401 5 2.22
7.6.1 It should be noted that the critical values in Table 5 for
From Table 1, for n = 14, we find that a value as large as 2.22 the 1 % level of significance are smaller than those for the 5 %
would occur by chance more than 5 % of the time, so we level. So for this particular test, the calculated value is
should retain the value 1.01 in further calculations. The Dixon significant if it is less than the chosen critical value.
test criterion is: 7.6.2 Example 5—In a comparison of strength of various
r 22 5 ~ x 14 2 x 12! / ~ x 14 2 x 3 ! plastic materials, one characteristic studied was the percentage
5 ~ 1.01 2 0.48! / ~ 1.0110.24! elongation at break. Before comparison of the average elonga-
50.53/1.25
tion of the several materials, it was desirable to isolate for
further study any pieces of a given material which gave very
50.424
small elongation at breakage compared with the rest of the
From Table 2 for n = 14, we see that the 5 % critical value pieces in the sample. Ten measurements of percentage elonga-
for r22 is 0.546. Since our calculated value (0.424) is less than tion at break made on a material are: 3.73, 3.59, 3.94, 4.13,
the critical value, we also retain 1.01 by Dixon’s test, and no 3.04, 2.22, 3.23, 4.05, 4.11, and 2.02. See Fig. 3. Arranged in
further values would be tested in this sample. ascending order of magnitude, these measurements are: 2.02,
7.5 Criteria for Two or More Outliers on Opposite Sides of 2.22, 3.04, 3.23, 3.59, 3.73, 3.94, 4.05, 4.11, 4.13.
the Sample—For suspected observations on both the high and 7.6.2.1 The questionable readings are the two lowest, 2.02
low sides in the sample, and to deal with the situation in which and 2.22. We can test these two low readings simultaneously
some of k ≥ 2 suspected outliers are larger and some smaller by using the S1,22/S2 criterion of Table 5. For the above
than the remaining values in the sample, Tietjen and Moore (7) measurements:
suggest the following statistic. Let the sample values be x1, x2, n

x3, ..., xn. Compute the sample mean, x̄ , and the n absolute S 2 5 Σ ~ x i 2 x̄ ! 2 5 5.351
i51
residuals: n n
S 21,2 5 Σ ~ x 2 x̄ 1,2 ! 2 5 1.196, where x̄ 1,2 5 Σ x i ⁄ ~ n 2 2 !
? ? ? ?
r 1 5 x 1 2 x̄ , r 2 5 x 2 2 x̄ , … , r n 5 x n 2 x̄ ? ? (10) i53 i53

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

5
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
TABLE 4 Tietjen-Moore Critical Values (One-Sided Test) for Ek
k 1A 2 3 4 5
n α 10 % 5% 1% 10 % 5% 1% 10 % 5% 1% 10 % 5% 1% 10 % 5% 1%
3 0.003 0.001 0.000 ... ... ... ... ... ... ... ... ... ... ... ...
4 0.049 0.025 0.004 0.002 0.001 0.000 ... ... ... ... ... ... ... ... ...
5 0.127 0.081 0.029 0.022 0.010 0.002 ... ... ... ... ... ... ... ... ...
6 0.203 0.145 0.068 0.056 0.034 0.012 0.009 0.004 0.001 ... ... ... ... ... ...
7 0.270 0.207 0.110 0.094 0.065 0.028 0.027 0.016 0.006 ... ... ... ... ... ...
8 0.326 0.262 0.156 0.137 0.099 0.050 0.053 0.034 0.014 0.016 0.010 0.004 ... ... ...
9 0.374 0.310 0.197 0.175 0.137 0.078 0.080 0.057 0.026 0.032 0.021 0.009 ... ... ...
10 0.415 0.353 0.235 0.214 0.172 0.101 0.108 0.083 0.044 0.052 0.037 0.018 0.022 0.014 0.006
11 0.451 0.390 0.274 0.250 0.204 0.134 0.138 0.107 0.064 0.073 0.055 0.030 0.036 0.026 0.012
12 0.482 0.423 0.311 0.278 0.234 0.159 0.162 0.133 0.083 0.094 0.073 0.042 0.052 0.039 0.020
13 0.510 0.453 0.337 0.309 0.262 0.181 0.189 0.156 0.103 0.116 0.092 0.056 0.068 0.053 0.031
14 0.534 0.479 0.374 0.337 0.293 0.207 0.216 0.179 0.123 0.138 0.112 0.072 0.086 0.068 0.042
15 0.556 0.503 0.404 0.360 0.317 0.238 0.240 0.206 0.146 0.160 0.134 0.090 0.105 0.084 0.054
16 0.576 0.525 0.422 0.384 0.340 0.263 0.263 0.227 0.166 0.182 0.153 0.107 0.122 0.102 0.068
17 0.593 0.544 0.440 0.406 0.362 0.290 0.284 0.248 0.188 0.198 0.170 0.122 0.140 0.116 0.079
18 0.610 0.562 0.459 0.424 0.382 0.306 0.304 0.267 0.206 0.217 0.187 0.141 0.156 0.132 0.094
19 0.624 0.579 0.484 0.442 0.398 0.323 0.322 0.287 0.219 0.234 0.203 0.156 0.172 0.146 0.108
20 0.638 0.594 0.499 0.460 0.416 0.339 0.338 0.302 0.236 0.252 0.221 0.170 0.188 0.163 0.121
25 0.692 0.654 0.571 0.528 0.493 0.418 0.417 0.381 0.320 0.331 0.298 0.245 0.264 0.236 0.188
30 0.730 0.698 0.624 0.582 0.549 0.482 0.475 0.443 0.386 0.391 0.364 0.308 0.325 0.298 0.250
35 0.762 0.732 0.669 0.624 0.596 0.533 0.523 0.495 0.435 0.443 0.417 0.364 0.379 0.351 0.299
40 0.784 0.756 0.704 0.657 0.629 0.574 0.562 0.534 0.480 0.486 0.458 0.408 0.422 0.395 0.347
45 0.802 0.776 0.728 0.684 0.658 0.607 0.593 0.567 0.518 0.522 0.492 0.446 0.459 0.433 0.386
50 0.820 0.796 0.748 0.708 0.684 0.636 0.622 0.599 0.550 0.552 0.529 0.482 0.492 0.468 0.424
A
From Grubbs (8),Table 1, for n # 25.

TABLE 5 Critical Values for S2n− 1, n / S2, or S21,2 / S2 for

Simultaneously Testing the Two Largest or Two Smallest
ObservationsA
FIG. 3 Ten Measurements of Percentage Elongation at Break
Lower 10 % Lower 5 % Lower 1 %
Number of from Example 5
Significance Significance Significance
Observations, n
Level Level Level
4 0.0031 0.0008 0.0000
5 0.0376 0.0183 0.0035
6 0.0920 0.0564 0.0186
critical value, we should conclude that both 2.02 and 2.22 are
7 0.1479 0.1020 0.0440 outliers. In a situation such as the one described in this
8 0.1994 0.1478 0.0750 example, where the outliers are to be isolated for further
9 0.2454 0.1909 0.1082
10 0.2863 0.2305 0.1414
analysis, a significance level as high as 5 % or perhaps even 10
11 0.3227 0.2667 0.1736 % would probably be used in order to get a reasonable size of
12 0.3552 0.2996 0.2043 sample for additional study.
13 0.3843 0.3295 0.2333
14 0.4106 0.3568 0.2605
7.6.3 Example 6—The following ranges (horizontal dis-
15 0.4345 0.3818 0.2859 tances in yards from gun muzzle to point of impact of a
16 0.4562 0.4048 0.3098 projectile) were obtained in firings from a weapon at a constant
17 0.4761 0.4259 0.3321
18 0.4944 0.4455 0.3530
angle of elevation and at the same weight of charge of
19 0.5113 0.4636 0.3725 propellant powder. The distances arranged in increasing order
20 0.5270 0.4804 0.3909 of magnitude are:
21 0.5415 0.4961 0.4082
22 0.5550 0.5107 0.4245 4420 4782
23 0.5677 0.5244 0.4398 4549 4803
24 0.5795 0.5373 0.4543 4730 4833
25 0.5906 0.5495 0.4680 4765 4838
26 0.6011 0.5609 0.4810
7.6.3.1 It is desired to make a judgment on whether the
27 0.6110 0.5717 0.4933
28 0.6203 0.5819 0.5050 projectiles exhibit uniformity in ballistic behavior or if some of
29 0.6292 0.5916 0.5162 the ranges are inconsistent with the others. The doubtful values
30 0.6375 0.6008 0.5268
35 0.6737 0.6405 0.5730
are the two smallest ranges, 4420 and 4549. For testing these
40 0.7025 0.6724 0.6104 two suspected outliers, the statistic S1,22/S2 is used. The value
45 0.7261 0.6985 0.6412 of S2 is 158592. Omission of the two shortest ranges, 4420 and
50 0.7459 0.7203 0.6672
4549, and recalculation, gives S1,22 equal to 8590.8. Thus:
A
From Grubbs (1), Table II. An observed ratio less than the appropriate critical
ratio in this table calls for rejection of the null hypothesis. S 1,2 ⁄S 2 5 8590.8⁄158592 5 0.0542

which is significant at the 0.01 level (see Table 5). It is thus

highly unlikely that the two shortest ranges (occurring actually
S 21,2 ⁄S 2 5 1.197⁄5.351 5 0.2237
from excessive yaw) could have come from the same popula-
From Table 5 for n = 10, the 5 % significance level for tion as that represented by the other six ranges. It should be
S1,22/S2 is 0.2305. Since the calculated value is less than the noted that the critical values in Table 5 for the 1 % level of

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

6
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
significance are smaller than those for the 5 % level. So for this should be used to test against change in level of several
particular test, the calculated value is significant if it is less observations in the same direction, and the sample coefficient
than the chosen critical value. of kurtosis:
NOTE 2—Kudo (10) indicates that if the two outliers are due to a shift n ~ n 1 1 ! Σ ~ x i 2 x̄ ! 4 3~n 2 1!2
in location or level, as compared to the scale σ, then the optimum sample g2 5 4 2 (16)
~ n 2 1 !~ n 2 2 !~ n 2 3 ! s ~ n 2 2 !~ n 2 3 !
criterion for testing should be of the type:
min (2 – xi – xj)/s = (2 – x1 – x2)/s in Example 5. is recommended to test against change in level to both higher
and lower values and also for changes in scale (variance).
7.7 Criteria for Two or More Outliers on the Same Side of
7.8.1 In applying the above tests, g1 or g2, or both, are
the Sample—An extension of the S 21,2 ⁄S 2 criterion is given by
computed and if their observed values exceed those for
Tietjen and Moore (7). Percentage points for the k ≥ 2 highest
significance levels given in Tables 7 and 8, then the observa-
or lowest sample values are given in Table 6, where:
tion farthest from the mean is rejected and the same procedure
n2k n
repeated until no further sample values are judged as outliers.
Lk 5 (
i51
~ x i 2 x̄ k ! 2 / ( ~ x i 2 x̄ ! 2
i51
(13)
Critical values in Tables 7 and 8 were obtained by simulation.
and: 7.8.2 Ferguson (11, 12) studied the power of the various
rejection rules relative to changes in level or scale. The g1
n2k
statistic has the optimum property of being “locally” best
x̄ k 5 ( x /~n 2 k!
i51
i (14)
against an alternative of shift in level (or mean) in the same
NOTE 3—For k = 1, L1 is equivalent to the statistic Tn for a single
outlier. For k = 2, L2 equals S n, n21 2 ⁄S 2 .
direction for multiple observations. g2 is similarly locally best
against alternatives of shift in both directions, or a of a change
7.8 Skewness and Kurtosis Criteria—When several outliers in scale for several observations. The g1 test is good for up to
are present in the sample, the detection of one or two spurious 50 % spurious observations in the sample for the one-sided
values may be “masked” by the presence of other anomalous case, and the g2 test is optimum in the two-sided alternatives
observations. So far we have discussed procedures for detect- case for up to 21 % “contamination” of sample values. For only
ing a fixed number of outliers in the same sample, but these one or two outliers the sample statistics of the previous
techniques are not generally the most sensitive. Sample skew- paragraphs are recommended, and Ferguson (11) discusses in
ness and kurtosis are defined in Practice E2586. They are detail their optimum properties of pointing out one or two
commonly used to test normality of a distribution, but may also outliers.
be used as outlier tests. Outlying observations occur due to a 7.8.3 Example 7—For the elongation at break data (Ex-
shift in level (or mean), or a change in scale (that is, change in ample 5), the value of skewness is g1 = –0.969. From Table 7
variance of the observations), or both. For several outliers and with n = 10, and taking into account that the two lowest values
repeated rejection of observations, the sample coefficient of are the suspected outliers, the 5 % significance value is –1.131,
skewness: with skewness less than this value being significant. The
nΣ ~ x i 2 x̄ ! 3 skewness test does not conclude that there are outliers in this
g1 5 (15)
~ n 2 1 !~ n 2 2 ! s 3 case.

TABLE 6 Tietjen-Moore Critical Values (One-Sided Test) for Lk

k 1A 2B 3 4 5
n α 10 % 5% 1% 10 % 5% 1% 10 % 5% 1% 10 % 5% 1% 10 % 5% 1%
3 0.011 0.003 0.000 ... ... ... ... ... ... ... ... ... ... ... ...
4 0.098 0.049 0.010 0.003 0.001 0.000 ... ... ... ... ... ... ... ... ...
5 0.199 0.127 0.044 0.038 0.018 0.004 ... ... ... ... ... ... ... ... ...
6 0.283 0.203 0.093 0.092 0.056 0.019 0.020 0.010 0.002 ... ... ... ... ... ...
7 0.350 0.270 0.145 0.148 0.102 0.044 0.056 0.032 0.010 ... ... ... ... ... ...
8 0.405 0.326 0.195 0.199 0.148 0.075 0.095 0.064 0.028 0.038 0.022 0.008 ... ... ...
9 0.450 0.374 0.241 0.245 0.191 0.108 0.134 0.099 0.048 0.068 0.045 0.018 ... ... ...
10 0.488 0.415 0.283 0.286 0.230 0.141 0.170 0.129 0.070 0.098 0.070 0.032 0.051 0.034 0.012
11 0.520 0.451 0.321 0.323 0.267 0.174 0.208 0.162 0.098 0.128 0.098 0.052 0.074 0.054 0.026
12 0.548 0.482 0.355 0.355 0.300 0.204 0.240 0.196 0.120 0.159 0.125 0.070 0.103 0.076 0.038
13 0.573 0.510 0.386 0.384 0.330 0.233 0.270 0.224 0.147 0.186 0.150 0.094 0.126 0.098 0.056
14 0.594 0.534 0.414 0.411 0.357 0.261 0.298 0.250 0.172 0.212 0.174 0.113 0.150 0.122 0.072
15 0.613 0.556 0.440 0.435 0.382 0.286 0.322 0.276 0.194 0.236 0.197 0.132 0.172 0.140 0.090
16 0.631 0.576 0.463 0.456 0.405 0.310 0.342 0.300 0.219 0.260 0.219 0.151 0.194 0.159 0.108
17 0.646 0.593 0.485 0.476 0.426 0.332 0.364 0.322 0.237 0.282 0.240 0.171 0.216 0.181 0.126
18 0.660 0.610 0.504 0.494 0.446 0.353 0.384 0.337 0.260 0.302 0.259 0.192 0.236 0.200 0.140
19 0.673 0.624 0.522 0.511 0.464 0.373 0.398 0.354 0.272 0.316 0.277 0.211 0.251 0.217 0.154
20 0.685 0.638 0.539 0.527 0.480 0.391 0.420 0.377 0.300 0.339 0.299 0.231 0.273 0.238 0.175
25 0.732 0.692 0.607 0.591 0.550 0.468 0.489 0.450 0.377 0.412 0.374 0.308 0.350 0.312 0.246
30 0.766 0.730 0.650 0.637 0.601 0.527 0.523 0.506 0.434 0.472 0.434 0.369 0.411 0.376 0.312
35 0.792 0.762 0.690 0.674 0.641 0.573 0.586 0.554 0.484 0.516 0.482 0.418 0.458 0.424 0.364
40 0.812 0.784 0.722 0.702 0.673 0.610 0.622 0.588 0.522 0.554 0.523 0.460 0.499 0.468 0.408
45 0.826 0.802 0.745 0.726 0.698 0.641 0.648 0.618 0.558 0.586 0.556 0.498 0.533 0.502 0.444
50 0.840 0.820 0.768 0.746 0.720 0.667 0.673 0.646 0.592 0.614 0.588 0.531 0.562 0.535 0.483
A
From Grubbs (8), Table I for n# 25.
B
From Grubbs (1), Table II.

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

7
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
TABLE 7 Significance LevelsA (One-Sided Test) for Skewness g1 TABLE 8 Significance LevelsA for Kurtosis g2
Number of 10 % 5% 1% Number of 10 % 5% 1%
Observations, Significance Significance Significance Observations, Significance Significance Significance
n Level Level Level n Level Level Level
3 1.647 1.711 1.731 4 3.075 3.518 3.900
4 1.439 1.709 1.940 5 2.772 3.506 4.454
5 1.224 1.564 1.994 6 2.482 3.319 4.685
6 1.090 1.428 1.959 7 2.257 3.110 4.735
7 1.014 1.320 1.886 8 2.067 2.935 4.687
8 0.956 1.246 1.813 9 1.904 2.772 4.586
9 0.903 1.183 1.735 10 1.778 2.627 4.467
10 0.862 1.131 1.668 11 1.678 2.505 4.350
11 0.828 1.086 1.610 12 1.597 2.399 4.234
12 0.798 1.049 1.556 13 1.529 2.300 4.106
13 0.770 1.011 1.504 14 1.471 2.217 4.000
14 0.744 0.977 1.461 15 1.422 2.145 3.887
15 0.722 0.950 1.418 16 1.378 2.081 3.784
16 0.702 0.922 1.379 17 1.340 2.021 3.702
17 0.684 0.899 1.345 18 1.303 1.966 3.605
18 0.667 0.875 1.310 19 1.271 1.921 3.524
19 0.651 0.856 1.281 20 1.243 1.873 3.450
20 0.636 0.836 1.252 21 1.214 1.831 3.370
21 0.624 0.818 1.225 22 1.188 1.788 3.298
22 0.610 0.800 1.196 23 1.167 1.757 3.233
23 0.599 0.786 1.175 24 1.143 1.719 3.169
24 0.587 0.770 1.150 25 1.123 1.690 3.116
25 0.578 0.757 1.132 26 1.102 1.658 3.051
26 0.567 0.743 1.108 27 1.085 1.630 2.995
27 0.558 0.731 1.091 28 1.066 1.601 2.943
28 0.549 0.718 1.070 29 1.052 1.578 2.903
29 0.541 0.708 1.056 30 1.035 1.550 2.845
30 0.532 0.695 1.036 35 0.969 1.446 2.642
35 0.497 0.649 0.965 40 0.913 1.358 2.470
40 0.467 0.610 0.904 45 0.867 1.285 2.322
45 0.442 0.578 0.853 50 0.830 1.223 2.210
50 0.422 0.551 0.812 A
Each entry calculated by 50 000 000 simulations.
A
Each entry calculated by 50 000 000 simulations.

estimate of standard deviation σ and n indicates the number of

7.8.4 Example 8—The kurtosis test is applied to the Venus observations in the sample under study.
semidiameter residuals data of Example 3 to test the highest 8.3 A slight over-approximation to critical values of T1' and
and lowest values. The value of kurtosis for the 15 observations Tn' is based on the Student’s t distribution:
is g2 = 2.528. The 5 % significance value from Table 8 is 2.145.
Using this test, we conclude that at least one of the values is an T 'n ~ α ! # t α⁄n,v =1 2 1⁄n (19)
outlier. With the value on the low side, –1.40, removed, the where tα/n,v is the upper α/n percentage point of Student’s t
value of skewness is g1 = 0.767. The 5 % significance value distribution with v degrees of freedom.
from Table 7 is 0.977, so no further outliers are concluded.
8.4 The population standard deviation σ may be known
8. Recommended Criterion Using an Independent accurately. In such cases, Table 10 may be used for single
Standard Deviation outliers.
8.1 Suppose that an independent estimate of the standard 9. Additional Comments: Reinforcement and New Issues
deviation is available from previous data. This estimate may be
9.1 The presence or lack of outliers is determined using
from a single sample of previous similar data or may be the
statistical testing on the basis of an underlying assumed normal
result of combining estimates from several such previous sets
distribution in this practice. Some additional remarks and
of data. When one uses an independent estimate of the standard
alternative approaches are noted.
deviation, sv, the test criterion for an outlier is as follows:
T' 1 5 ~ x̄ 2 x 1 ! /s v (17)
9.2 If the mathematical form of the underlying uncontami-
nated statistical distribution is known and not normal or
or: transformable to normal, for example, an exponential life
T' n 5 ~ x n 2 x̄ ! /s v (18) distribution, then outlier testing should specifically account for
it. Some classes of data provide distributions that are highly
where: asymmetric (skewed).
v = total number of degrees of freedom. 9.3 In general, the more is known about data variation, the
8.2 Critical values for T1' and Tn' given by David (13) are in better a position the experimenter is in to test for outliers.
Table 9. In Table 9 the subscript v = df indicates the total Outlier tests provided can be classified based on availability of
number of degrees of freedom associated with the independent prior information on variation: nothing known (Tables 1 and

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

8
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
TABLE 9 Critical Values (One-Sided Test) for T' When Standard Deviation s v is Independent of Present SampleA
x n 2 x̄ x̄ 2 x 1
T' 5 , or
sv sv
n
v = d.f.
3 4 5 6 7 8 9 10 12
1 % significance level
10 2.78 3.10 3.32 3.48 3.62 3.73 3.82 3.90 4.04
11 2.72 3.02 3.24 3.39 3.52 3.63 3.72 3.79 3.93
12 2.67 2.96 3.17 3.32 3.45 3.55 3.64 3.71 3.84
13 2.63 2.92 3.12 3.27 3.38 3.48 3.57 3.64 3.76
14 2.60 2.88 3.07 3.22 3.33 3.43 3.51 3.58 3.70
15 2.57 2.84 3.03 3.17 3.29 3.38 3.46 3.53 3.65
16 2.54 2.81 3.00 3.14 3.25 3.34 3.42 3.49 3.60
17 2.52 2.79 2.97 3.11 3.22 3.31 3.38 3.45 3.56
18 2.50 2.77 2.95 3.08 3.19 3.28 3.35 3.42 3.53
19 2.49 2.75 2.93 3.06 3.16 3.25 3.33 3.39 3.50
20 2.47 2.73 2.91 3.04 3.14 3.23 3.30 3.37 3.47
24 2.42 2.68 2.84 2.97 3.07 3.16 3.23 3.29 3.38
30 2.38 2.62 2.79 2.91 3.01 3.08 3.15 3.21 3.30
40 2.34 2.57 2.73 2.85 2.94 3.02 3.08 3.13 3.22
60 2.29 2.52 2.68 2.79 2.88 2.95 3.01 3.06 3.15
120 2.25 2.48 2.62 2.73 2.82 2.89 2.95 3.00 3.08
` 2.22 2.43 2.57 2.68 2.76 2.83 2.88 2.93 3.01
5 % significance level
10 2.01 2.27 2.46 2.60 2.72 2.81 2.89 2.96 3.08
11 1.98 2.24 2.42 2.56 2.67 2.76 2.84 2.91 3.03
12 1.96 2.21 2.39 2.52 2.63 2.72 2.80 2.87 2.98
13 1.94 2.19 2.36 2.50 2.60 2.69 2.76 2.83 2.94
14 1.93 2.17 2.34 2.47 2.57 2.66 2.74 2.80 2.91
15 1.91 2.15 2.32 2.45 2.55 2.64 2.71 2.77 2.88
16 1.90 2.14 2.31 2.43 2.53 2.62 2.69 2.75 2.86
17 1.89 2.13 2.29 2.42 2.52 2.60 2.67 2.73 2.84
18 1.88 2.11 2.28 2.40 2.50 2.58 2.65 2.71 2.82
19 1.87 2.11 2.27 2.39 2.49 2.57 2.64 2.70 2.80
20 1.87 2.10 2.26 2.38 2.47 2.56 2.63 2.68 2.78
24 1.84 2.07 2.23 2.34 2.44 2.52 2.58 2.64 2.74
30 1.82 2.04 2.20 2.31 2.40 2.48 2.54 2.60 2.69
40 1.80 2.02 2.17 2.28 2.37 2.44 2.50 2.56 2.65
60 1.78 1.99 2.14 2.25 2.33 2.41 2.47 2.52 2.61
120 1.76 1.96 2.11 2.22 2.30 2.37 2.43 2.48 2.57
` 1.74 1.94 2.08 2.18 2.27 2.33 2.39 2.44 2.52
10 % significance level
10 1.68 1.92 2.09 2.23 2.33 2.42 2.50 2.56 2.68
11 1.66 1.90 2.07 2.20 2.30 2.39 2.46 2.53 2.64
12 1.65 1.88 2.05 2.17 2.28 2.36 2.44 2.50 2.61
13 1.63 1.86 2.03 2.16 2.26 2.34 2.41 2.47 2.58
14 1.62 1.85 2.01 2.14 2.24 2.32 2.39 2.45 2.56
15 1.61 1.84 2.00 2.12 2.22 2.31 2.38 2.44 2.54
16 1.61 1.83 1.99 2.11 2.21 2.29 2.36 2.42 2.52
17 1.60 1.82 1.98 2.10 2.20 2.28 2.35 2.41 2.51
18 1.59 1.82 1.97 2.09 2.19 2.27 2.34 2.39 2.49
19 1.59 1.81 1.96 2.08 2.18 2.26 2.33 2.38 2.48
20 1.58 1.80 1.96 2.08 2.17 2.25 2.32 2.37 2.47
24 1.57 1.78 1.94 2.05 2.15 2.22 2.29 2.34 2.44
30 1.55 1.77 1.92 2.03 2.12 2.20 2.26 2.32 2.41
40 1.54 1.75 1.90 2.01 2.10 2.17 2.23 2.29 2.38
60 1.52 1.73 1.87 1.98 2.07 2.14 2.20 2.26 2.35
120 1.51 1.71 1.85 1.96 2.05 2.12 2.18 2.23 2.32
` 1.50 1.70 1.83 1.94 2.02 2.09 2.15 2.20 2.28
A
The percentage points are reproduced from Ref. (13).

2), limited historical information (Table 9), standard deviation 9.5 All of the documented test methodologies are univari-
known (Table 10). A cautionary note is that a historical ate. This practice does not address the issue of multivariate
variation estimate must still be relevant. outlier testing or testing in time-ordered or structured data.
9.4 Much outlier practice is directed towards a more reliable 9.6 The outlier tests provided in this practice are generally
estimate of a measure of the mean. If a goal of study is instead most useful with moderate numbers of observations. Outlier
to make inferences about variability or to estimate a relatively tests that only use information about variability internal to the
low or high quantile of the distribution, then any action that is sample can only reject gross outlying values. With much larger
taken with the disposition of perceived outliers dramatically numbers of observations, especially in data sets that have not
changes the resulting statistical estimates and interpretation. been screened by a knowledgeable reviewer to remove invalid

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

9
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
TABLE 10 Critical ValuesA (One-Sided Test) of T'1` and T'n` When are the Tukey boxplot rule and Hampel’s Rule. The commonly
the Population Standard Deviation σ is Known
used rejection criteria for each rule are still selected to provide
Number of 10 % 5% 1%
Observations, Significance Significance Significance
a reasonable significance level(s) for an assumed underlying
n Level Level Level uncontaminated normal distribution.
2 1.163 1.386 1.822 9.7.1 The Tukey boxplot rule identifies as potential outliers
3 1.497 1.737 2.216
4 1.696 1.941 2.431 values more than a multiple (1.5) of the interquartile range
5 1.834 2.080 2.574 from the lower or upper quartile of a data set (Practice E2586),
6 1.939 2.184 2.679 and rejects as probable outliers values more than 3.0 times the
7 2.022 2.266 2.761
8 2.091 2.334 2.827 interquartile range from the lower or upper quartile. The first
9 2.149 2.392 2.884 (1.5) rule, on average, will identify approximately 1% of data
10 2.200 2.441 2.932 originating from a normal distribution as potential outliers.
11 2.245 2.485 2.973
12 2.284 2.523 3.009 Robustness with multiple outliers is achieved because quartiles
13 2.320 2.558 3.042 are little influenced by outlier values.
14 2.352 2.589 3.072 9.7.2 Hampel’s rule rejects values that are farther than a
15 2.382 2.618 3.099
16 2.409 2.644 3.124 multiple (4.5 or 5.2) of the median absolute deviation away
17 2.434 2.668 3.147 from the median of the data set. It achieves its robustness
18 2.458 2.691 3.168 through use of both the median and median absolute deviation
19 2.480 2.712 3.187
20 2.500 2.732 3.206 to reduce the impact that outliers have on the test statistic.
21 2.520 2.750 3.223 Multiples 4.5 and 5.2 median absolute deviations correspond to
22 2.538 2.768 3.240 approximately 3.0 and 3.5 standard deviations from the center
23 2.556 2.785 3.255
24 2.572 2.800 3.270 for a normal distribution.
25 2.588 2.815 3.284 9.7.3 The alternative procedures also apply to non-normal
26 2.602 2.829 3.297 distributions, but do implicitly assume symmetry. Robustness
27 2.617 2.844 3.310
28 2.631 2.857 3.322 breaks down in the presence of a meaningfully asymmetric
29 2.644 2.869 3.334 non-normal underlying distribution.
30 2.656 2.881 3.345
35 2.712 2.935 3.395 9.8 Outlier Accommodation—Robust statistical methods are
40 2.760 2.980 3.437 insensitive to small numbers of outlier data. Examples are use
45 2.801 3.019 3.472
50 2.837 3.054 3.504
of the median or trimmed mean as estimates of the mean, and
A
least absolute deviations for regression. Many robust estima-
Each entry calculated by 20 000 000 simulations.
tion methods have been developed, but have not yet gained the
wide use to be considered standard replacements for the
observations, the presence of invalid data is to be expected. customary least squares methods.
The statistical basis for the tests in the previous sections, that 9.9 Additional literature and monographs that summarize a
there should be a low probability of rejecting any value if the range of viewpoints on the detection and handling of outliers
distribution is normal, is less compelling in that case. are listed in Refs. (9, 11, 14-19).
9.7 Alternative Outlier Procedures—Outlier rejection rules
based on robust statistical measures have been introduced. A 10. Keywords
purpose of the robust methodology is to reduce the impact of 10.1 Dixon test; gross deviation; Grubbs test; kurtosis;
multiple outliers on the test statistic. Two common examples outlier; skewness; Tietjen-Moore test

REFERENCES

(1) Grubbs, F. E., and Beck, G., “Extension of Sample Sizes and (6) Chauvenet, W., Method of Least Squares, Lippincott, Philadelphia,
Percentage Points for Significance Tests of Outlying Observations,” 1868.
Technometrics, TCMTA, Vol 14, No. 4, November 1972, pp. 847–854. (7) Tietjen, G. L., and Moore, R. H., “Some Grubbs-Type Statistics for
(2) Dixon, W. J., “Processing Data for Outliers,” Biometrics, BIOMA, Vol the Detection of Several Outliers,” Technometrics, TCMTA, Vol 14,
9, No. 1, March 1953, pp. 74–89. No. 3, August 1972, pp. 583–597. Corrigendum Technometrics, Vol
(3) Bohrer, A., “One-sided and Two-sided Critical Values for Dixon’s 21, No. 3, August 1979, p. 396.
Outlier Test for Sample Sizes up to n=30,” Economic Quality Control, (8) Grubbs, F. E., “Sample Criteria for Testing Outlying Observations,”
Vol 23, No. 1, 2008, pp. 5–13. Annals of Mathematical Statistics, AASTA, Vol 21, March 1950, pp.
(4) Verma, S. P., and Quiroz-Ruiz, A., “Critical Values for Six Dixon 27–58.
Tests for Outliers in Normal Samples up to Sizes 100, and Applica- (9) Grubbs, F. E., “Procedures for Detecting Outlying Observations in
tions in Science and Engineering,” Revista Mexicana de Ciencias Samples,” Technometrics, TCMTA, Vol 11, No. 4, February 1969, pp.
Geologicas, Vol 23, No. 2, 2006, pp. 133–161. 1–21.
(5) David, H. A., Hartley, H. O., and Pearson, E. S., “The Distribution of (10) Kudo, A., “On the Testing of Outlying Observations,” Sankhya, The
the Ratio, in a Single Normal Sample, of Range to Standard Indian Journal of Statistics, SNKYA, Vol 17, Part 1, June 1956, pp.
Deviation,” Biometrika, BIOKA, Vol 41, 1954, pp. 482–493. 67–76.

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

10
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
 E178 − 21
(11) Ferguson, T. S., “On the Rejection of Outliers,” Fourth Berkeley (15) Barnett, V., “The Study of Outliers: Purpose and Model,” Applied
Symposium on Mathematical Statistics and Probability, edited by Statistics, Vol 27, 1978, pp. 242–250.
Jerzy Neyman, University of California Press, Berkeley and Los (16) Hawkins, D. M., Identification of Outliers, Chapman and Hall,
Angeles, Calif., 1961. London, 1980.
(12) Ferguson, T. S., “Rules for Rejection of Outliers,” Revue Inst. Int. de (17) Beckman, R. J., and Cook, R. D., “Outlier……….s,” Technometrics,
Stat., RINSA, Vol 29, No. 3, 1961, pp. 29–43. Vol 25, No. 2, 1983, pp. 119–149.
(13) David, H. A., “Revised Upper Percentage Points of the Extreme (18) Iglewicz, B., and Hoaglin, D. C., How to Detect and Handle
Studentized Deviate from the Sample Mean,” Biometrika, BIOKA, Outliers, ASQ Quality Press, 1993.
Vol 43, 1956, pp. 449–451. (19) Barnett, V. and Lewis, T., Outliers in Statistical Data, 3rd ed., John
(14) Anscombe, F. J.,“Rejection of Outliers,” Technometrics, TCMTA,
Wiley and Sons, Inc., New York, 1995.
Vol 2, No. 2, 1960, pp. 123–147.

ASTM International takes no position respecting the validity of any patent rights asserted in connection with any item mentioned
in this standard. Users of this standard are expressly advised that determination of the validity of any such patent rights, and the risk
of infringement of such rights, are entirely their own responsibility.

This standard is subject to revision at any time by the responsible technical committee and must be reviewed every five years and
if not revised, either reapproved or withdrawn. Your comments are invited either for revision of this standard or for additional standards
and should be addressed to ASTM International Headquarters. Your comments will receive careful consideration at a meeting of the
responsible technical committee, which you may attend. If you feel that your comments have not received a fair hearing you should
make your views known to the ASTM Committee on Standards, at the address shown below.

This standard is copyrighted by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959,
United States. Individual reprints (single or multiple copies) of this standard may be obtained by contacting ASTM at the above
address or at 610-832-9585 (phone), 610-832-9555 (fax), or [email protected] (e-mail); or through the ASTM website
(www.astm.org). Permission rights to photocopy the standard may also be secured from the Copyright Clearance Center, 222
Rosewood Drive, Danvers, MA 01923, Tel: (978) 646-2600; http://www.copyright.com/

&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG 0RQ$SU87&

11
'RZQORDGHGSULQWHGE\
2KLR8QLYHUVLW\SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG