Lec 5 - Normality Testing

Testing for Normality
For each mean and standard deviation combination a theoretical

normal distribution can be determined. This distribution is based
on the proportions shown below.
This theoretical normal distribution can then be compared to the

actual distribution of the data.
The actual data
distribution that has a
mean of 66.51 and a
standard deviation of
18.265.
Theoretical normal
distribution calculated
from a mean of 66.51 and a
standard deviation of
18.265.
Are the actual data statistically different than the computed

normal curve?
There are several methods of assessing whether data are

normally distributed or not. They fall into two broad categories:
graphical and statistical. The some common techniques are:
Graphical
Q-Q probability plots
Cumulative frequency (P-P) plots
Statistical
W/S test
Jarque-Bera test
Shapiro-Wilks test
Kolmogorov-Smirnov test
DAgostino test
Q-Q plots display the observed values against normally

distributed data (represented by the line).
Normally distributed data fall along the line.
Graphical methods are typically not very useful when the sample size is
small. This is a histogram of the last example. These data do not look
normal, but they are not statistically different than normal.
Tests of Normality
a
Age
Kolmogorov-Smirnov
df
Sig.
Statistic
1048
.000
.110
Shapiro-Wilk
df
Statistic
.931
1048
Sig.
.000
a. Lilliefors Significance Correction
Tests of Normality
a
TOTAL_VALU
Kolmogorov-Smirnov
Statistic
df
Sig.
.283
149
.000
Statistic
.463
Shapiro-Wilk
df
149
Sig.
.000
Tests of Normality
a
Z100
Kolmogorov-Smirnov
Statistic
df
Sig.
.071
100
.200*
Statistic
.985
*. This is a lower bound of the true s ignificance.

Shapiro-Wilk
df
100
Sig.
.333
Statistical tests for normality are more precise since actual

probabilities are calculated.
Tests for normality calculate the probability that the sample was
drawn from a normal population.
The hypotheses used are:
Ho: The sample data are not significantly different than a
normal population.
Ha: The sample data are significantly different than a normal
population.
Typically, we are interested in finding a difference between groups.

When we are, we look for small probabilities.
If the probability of finding an event is rare (less than 5%) and
we actually find it, that is of interest.
When testing normality, we are not looking for a difference.
In effect, we want our data set to be NO DIFFERENT than
normal. We want to accept the null hypothesis.
So when testing for normality:
Probabilities > 0.05 mean the data are normal.
Probabilities < 0.05 mean the data are NOT normal.
Remember that LARGE probabilities denote normally distributed

data. Below are examples taken from SPSS.
Normally Distributed Data
a
Asthma Cases
Kolmogorov-Smirnov
Statistic
df
Sig.
.069
.200*
72
Shapiro-Wilk
Statistic
df
72
.988
Sig.
.721
*. This is a lower bound of the true significance.

Non-Normally Distributed Data

a
Average PM10
Kolmogorov-Smirnov
Statistic
df
Sig.
.142
72
.001
Shapiro-Wilk
Statistic
df
.841
72
Sig.
.000
Normally Distributed Data

a
Asthma Cases
Kolmogorov-Smirnov
Statistic
df
Sig.
.069
.200*
72
Shapiro-Wilk
Statistic
df
72
.988
Sig.
.721
*. This is a lower bound of the true significance.

In SPSS output above the probabilities are greater than 0.05 (the
typical alpha level), so we accept Ho these data are not different
from normal.
Non-Normally Distributed Data

a
Average PM10
Kolmogorov-Smirnov
Statistic
df
Sig.
.142
72
.001
Shapiro-Wilk
Statistic
df
.841
72
Sig.
.000
In the SPSS output above the probabilities are less than 0.05 (the typical
alpha level), so we reject Ho these data are significantly different from
normal.
Important: As the sample size increases, normality parameters becomes
MORE restrictive and it becomes harder to declare that the data are
normally distributed. So for very large data sets, normality testing
becomes less important.
Three Simple Tests for Normality
W/S Test for Normality

A fairly simple test that requires only the sample standard
deviation and the data range.
Based on the q statistic, which is the studentized (meaning t
distribution) range, or the range expressed in standard
deviation units. Tests kurtosis.
Should not be confused with the Shapiro-Wilks test.
w
q=
s
where q is the test statistic, w is the range of the data and s is
the standard deviation.
Range constant,
SD changes
Range changes,
SD constant
Village
Aranza
Corupo
San Lorenzo
Cheranatzicurin
Nahuatzen
Pomacuaran
Sevina
Arantepacua
Cocucho
Charapan
Comachuen
Pichataro
Quinceo
Nurio
Turicuaro
Urapicho
Capacuaro
Population
Density
4.13
4.53
4.69
4.76
4.77
4.96
4.97
5.00
5.04
5.10
5.25
5.36
5.94
6.06
6.19
6.30
7.73
Standard deviation (s) = 0.866

Range (w) = 3.6
n = 17
q=
w
s
q=
3.6
= 4.16
0.866
qCritical Range = 3.06 to 4.31
The W/S test uses a critical range. IF the calculated value falls WITHIN the range,
then accept Ho. IF the calculated value falls outside the range then reject Ho.
Since 3.06 < q=4.16 < 4.31, then we accept Ho.
Since we have a critical range, it is difficult to determine a probability

range for our results. Therefore we simply state our alpha level.
The sample data set is not significantly different than normal
(W/S4.16, p > 0.05).
JarqueBera Test
A goodness-of-fit test of whether sample data have the skewness
and kurtosis matching a normal distribution.
n
k3 =
( xi x )
i =1
ns
k4 =
4
x
x
(
)
i
i =1
ns
(k 3 )2 (k 4 )2
+
JB = n
24
6
Where x is each observation, n is the sample size, s is the

standard deviation, k3 is skewness, and k4 is kurtosis.
x = 5.34
s = 0.87
Village
Aranza
Corupo
San Lorenzo
Cheranatzicurin
Nahuatzen
Pomacuaran
Sevina
Arantepacua
Cocucho
Charapan
Comachuen
Pichataro
Quinceo
Nurio
Turicuaro
Urapicho
Capacuaro
Population
Density
4.13
4.53
4.69
4.76
4.77
4.96
4.97
5.00
5.04
5.10
5.25
5.36
5.94
6.06
6.19
6.30
7.73
Mean
Deviates
-1.21
-0.81
-0.65
-0.58
-0.57
-0.38
-0.37
-0.34
-0.30
-0.24
-0.09
0.02
0.60
0.72
0.85
0.96
2.39
Mean
Deviates3
-1.771561
-0.531441
-0.274625
-0.195112
-0.185193
-0.054872
-0.050653
-0.039304
-0.027000
-0.013824
-0.000729
0.000008
0.216000
0.373248
0.614125
0.884736
13.651919
12.595722
Mean
Deviates4
2.14358881
0.43046721
0.17850625
0.11316496
0.10556001
0.02085136
0.01874161
0.01336336
0.00810000
0.00331776
0.00006561
0.00000016
0.12960000
0.26873856
0.52200625
0.84934656
32.62808641
37.433505
12.6
k3 =
= 1.13
3
(17)(0.87 )
37.43
k4 =
3 = 0.843
4
(17)(0.87 )
(1.13)2 (0.843)2
1.2769 0.711
= 17
+
JB = 17
+
24
24
6
6
JB = 17(0.2128 + 0.0296 )
JB = 4.12
The Jarque-Bera statistic can be compared to the 2 distribution (table) with 2

degrees of freedom (df or v) to determine the critical value at an alpha level of
0.05.
The critical 2 value is 5.991. Our calculated Jarque-Bera statistic is 4.12

which falls between 0.5 and 0.1, which is greater than the critical value.
Therefore we accept Ho that there is no difference between our distribution
and a normal distribution (Jarque-Bera 24.15, 0.5 > p > 0.1).
DAgostino Test
A very powerful test for departures from normality.
Based on the D statistic, which gives an upper and lower critical
value.
D=
T
3
n SS
where
n +1
T = i
Xi
2
where D is the test statistic, SS is the sum of squares of the data

and n is the sample size, and i is the order or rank of observation
X. The df for this test is n (sample size).
First the data are ordered from smallest to largest or largest to
smallest.
Village
Aranza
Corupo
San Lorenzo
Cheranatzicurin
Nahuatzen
Pomacuaran
Sevina
Arantepacua
Cocucho
Charapan
Comachuen
Pichataro
Quinceo
Nurio
Turicuaro
Urapicho
Capacuaro
Population
Density
4.13
4.53
4.69
4.76
4.77
4.96
4.97
5.00
5.04
5.10
5.25
5.36
5.94
6.06
6.19
6.30
7.73
Mean
i
Deviates2
1
1.46410
2
0.65610
3
0.42250
4
0.33640
5
0.32490
6
0.14440
7
0.13690
8
0.11560
9
0.09000
10
0.05760
11
0.00810
12
0.00040
13
0.36000
14
0.51840
15
0.72250
16
0.92160
17
5.71210
Mean = 5.34 SS = 11.9916
(4.13 5.34) 2 = 1.212 = 1.46410
n + 1 17 + 1
=
=9
2
2
T = (i 9) X 1
T = (1 9)4.13 + (2 9)4.53 + (3 9)4.69 + ...(17 9)7.73
T = 63.23
D=
63.23
(17 3 )(11.9916)
= 0.26050
DCritical = 0.2587,0.2860
Df = n = 17
If the calculated value falls within the critical range, accept Ho.
Since 0.2587 < D = 0.26050 < 0.2860 accept Ho.
The sample data set is not significantly different than normal (D0.26050, p > 0.05).
Use the next lower

n on the table if your
sample size is NOT
listed.
Different normality tests produce vastly different probabilities. This is due

to where in the distribution (central, tails) or what moment (skewness, kurtosis)
they are examining.
Normality Test
W/S
Jarque-Bera
DAgostino
Shapiro-Wilk
Kolmogorov-Smirnov
Statistic
q
2
D
W
D
Calculated Value
4.16
4.15
0.2605
0.8827
0.2007
Probability
> 0.05
0.5 > p > 0.1
> 0.05
0.035
0.067
Results
Normal
Normal
Normal
Non-normal
Normal
Normality tests using various random normal sample sizes:

Sample
Size
20
50
100
200
500
1000
2000
5000
10000
15000
20000
50000
K-S
Prob
1.000
0.893
0.871
0.611
0.969
0.904
0.510
0.106
0.007
0.005
0.007
0.002
Notice that as the sample size increases, the probabilities decrease.

In other words, it gets harder to meet the normality assumption as
the sample size increases since even small differences are detected.
Which normality test should I use?

Kolmogorov-Smirnov:
Not sensitive to problems in the tails.
For data sets > 50.
Shapiro-Wilks:
Doesn't work well if several values in the data set are the same.
Works best for data sets with < 50, but can be used with larger
data sets.
W/S:
Simple, but effective.
Not available in SPSS.
Jarque-Bera:
Tests for skewness and kurtosis, very effective.
DAgostino:
Powerful omnibus (skewness, kurtosis, centrality) test.
When is non-normality a problem?

Normality can be a problem when the sample size is small (< 50).
Highly skewed data create problems.
Highly leptokurtic data are problematic, but not as much as
skewed data.
Normality becomes a serious concern when there is activity in
the tails of the data set.
Outliers are a problem.
Clumps of data in the tails are worse.
Final Words Concerning Normality Testing:

1. Since it IS a test, state a null and alternate hypothesis.
2. If you perform a normality test, do not ignore the results.
3. If the data are not normal, use non-parametric tests.
4. If the data are normal, use parametric tests.
AND MOST IMPORTANTLY:
5. If you have groups of data, you MUST test each group for
normality.

Lec 5 - Normality Testing

Uploaded by

Copyright:

Available Formats

Lec 5 - Normality Testing

Uploaded by

Document Information

Original Description:

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

Lec 5 - Normality Testing

Uploaded by

Copyright:

Available Formats

Testing for Normality

For each mean and standard deviation combination a theoretical

This theoretical normal distribution can then be compared to the

Are the actual data statistically different than the computed

There are several methods of assessing whether data are

Q-Q plots display the observed values against normally

Normally distributed data fall along the line.

a. Lilliefors Significance Correction

a. Lilliefors Significance Correction

*. This is a lower bound of the true s ignificance.

Statistical tests for normality are more precise since actual

Typically, we are interested in finding a difference between groups.

Remember that LARGE probabilities denote normally distributed

*. This is a lower bound of the true significance.

Non-Normally Distributed Data

a. Lilliefors Significance Correction

Normally Distributed Data

*. This is a lower bound of the true significance.

Non-Normally Distributed Data

a. Lilliefors Significance Correction

Three Simple Tests for Normality

W/S Test for Normality

Standard deviation (s) = 0.866

qCritical Range = 3.06 to 4.31

Since we have a critical range, it is difficult to determine a probability

Where x is each observation, n is the sample size, s is the

The Jarque-Bera statistic can be compared to the 2 distribution (table) with 2

The critical 2 value is 5.991. Our calculated Jarque-Bera statistic is 4.12

where D is the test statistic, SS is the sum of squares of the data

(4.13 5.34) 2 = 1.212 = 1.46410

Use the next lower

Different normality tests produce vastly different probabilities. This is due

Normality tests using various random normal sample sizes:

Notice that as the sample size increases, the probabilities decrease.

Which normality test should I use?

When is non-normality a problem?

Final Words Concerning Normality Testing:

You might also like