TESTING DATA FOR NORMALITY

Introduction

In advanced statistics, one of the most important preliminary steps before selecting an appropriate statistical
test is determining whether the data follows a normal distribution. Many statistical techniques, especially
parametric tests such as t-tests and ANOVA, assume that the data are normally distributed. However, in real-
life educational, social, and behavioral data, this assumption is often violated. This is where testing for
normality becomes essential. By identifying whether the data deviate from normality, researchers can decide
whether to use parametric or nonparametric methods. Nonparametric tests are particularly useful when data
are skewed, ordinal, or contain outliers. Understanding normality also helps in interpreting the shape, spread,
and central tendency of data. Tools such as skewness, kurtosis, and statistical tests like the Kolmogorov-
Smirnov test are commonly used to assess normality. For future math educators, mastering these concepts
ensures proper analysis of classroom data such as test scores and survey results. This lesson will guide you
through both descriptive and inferential methods of testing normality. Ultimately, knowing when to shift to
nonparametric methods strengthens your statistical decision-making.

Objectives
At the end of this lesson, students should be able to:
1. Compute measures of normality by determining skewness and kurtosis using step-by-step procedures.
2. Perform and interpret results of normality tests to determine whether a dataset follows a normal
distribution.
3. Make statistical decisions based on normality results by selecting appropriate parametric or
nonparametric tests for given data sets.

Discussion

A normal distribution is a symmetrical, bell-shaped curve where most data cluster around the mean. It has the
following characteristics:

• Mean = Median = Mode

• Symmetrical shape
• Predictable spread using standard deviation

In contrast, non-normal distributions may be:

• Skewed (left or right)

• Flat or peaked (kurtosis differences)
In figures:

Examples

1. Scores of students in a well-designed national exam often approximate normal distribution.

2. Income levels of families are usually right-skewed.
3. Quiz scores in a difficult math test may be left-skewed.

COMPUTING AND TESTING KURTOSIS AND SKEWNESS FOR SAMPLE NORMALITY

The kurtosis and skewness can be used to determine if a sample approximately resembles a normal
distribution. There are five steps for examining sample normality in terms of kurtosis and skewness.

1. Determine the sample’s mean and standard deviation.

2. Determine the sample’s kurtosis and skewness.

3. Calculate the standard error of the kurtosis and the standard error of the skewness.

4. Calculate the z-score for the kurtosis and the z-score for the skewness.

5. Compare the z-scores with the critical region obtained from the normal distribution.

Compare these z-scores with the values of the normal distribution for a desired level of confidence.
For example, if you set alpha is 0.05, then the calculated z-scores for an approximately normal
distribution must fall between −1.96 and +1.96.
Sample Problem
The scores below represent students’ quiz performance during the first week of class. Use _ = 0.05 for your
desired level of confidence. Determine if the samples of week 1 quiz scores are approximately normal in terms
of its kurtosis.

1. Determine the sample’s mean and standard deviation.

∑𝑥 1706
𝑥̅ = = = 80.24
𝑛 21

∑(𝑥 − 𝑥̅ )2 5525.81
𝑠=√ = √ = √276.29 = 16.62
𝑛−1 21 − 1
2. Determine the sample’s kurtosis and skewness.

𝑛 (𝑛 + 1) 𝑥 − 𝑥̅ 4 3 (𝑛 + 1)2
𝐾= [ ∑( ) ]−
(𝑛 − 1)(𝑛 − 2)(𝑛 − 3) 𝑠 (𝑛 − 2)(𝑛 − 3)
21 (21 + 1) 3 (21 + 1)2
𝐾= [ (69.058)] − = 𝟏. 𝟏𝟓𝟔
(21 − 1)(21 − 2)(21 − 3) (21 − 2)(21 − 3)

𝑛 𝑥 − 𝑥̅ 3 21
𝑆𝑘 = [ ∑( ) ]=[ (−18.427)] = −𝟏. 𝟎𝟏𝟖
(𝑛 − 1)(𝑛 − 2) 𝑠 (21 − 1)(21 − 2)

3. Calculate the standard error of the kurtosis and the standard error of the skewness.
24𝑛 (𝑛 − 1)2 (24)(21) (21 − 1)2
𝑆𝐸𝑘 = √ =√ = 𝟎. 𝟗𝟕𝟐
(𝑛 − 2)(𝑛 − 3)(𝑛 + 5)(𝑛 + 3) (21 − 2)(21 − 3)(21 + 5)(21 + 3)

6𝑛 (𝑛 − 1) 6(21) (21 − 1)
𝑆𝐸𝑆𝑘 = √ =√ = 𝟎. 𝟓𝟎𝟏
(𝑛 − 2)(𝑛 + 1)(𝑛 + 3) (21 − 2)(21 + 1)(21 + 3)
4. Calculate the z-score for the kurtosis and the z-score for the skewness.

𝐾 − 0 1.156 − 0
𝑧𝑘 = = = 1.189
𝑆𝐸𝑘 0.972

𝑆𝑘 − 0 −1.018
𝑧𝑆𝑘 = = = −2.032
𝑆𝐸𝑆𝑘 0.501
5. Compare the z-scores with the critical region obtained from the normal distribution.

Use the z-score to examine the sample’s approximation to a normal distribution. This value must fall
between −1.96 and +1.96 to pass the normality assumption for α = 0.05. Since this z-score of 1.189
value does fall within that range, the sample has passed our normality assumption for kurtosis while z-
score -2.032 value does not fall within that range, the sample has failed our normality assumption for
skewness. Therefore, you must use a nonparametric statistical test.
Table for computations

𝑥 − 𝑥̅ 𝑥 − 𝑥̅ 4 𝑥 − 𝑥̅ 3
DATA (x) 𝑥 − 𝑥̅ (𝑥 − 𝑥̅ )2 ( ) ( )
𝑠 𝑠 𝑠
90 9.76 95.26 0.587 0.119 0.202
72 -8.24 67.90 -0.496 0.061 -0.122
90 9.76 95.26 0.587 0.119 0.202
64 -16.24 263.74 -0.977 0.911 -0.933
95 14.76 217.86 0.888 0.622 0.7
89 8.76 76.74 0.527 0.077 0.146
74 -6.24 38.94 -0.375 0.02 -0.053
88 7.76 60.22 0.467 0.048 0.102
100 19.76 390.46 1.189 1.999 1.681
77 -3.24 10.50 -0.195 0.001 -0.007
57 -23.24 540.10 -1.398 3.82 -2.732
35 -45.24 2046.66 -2.722 54.897 -20.168
100 19.76 390.46 1.189 1.999 1.681
64 -16.24 263.74 -0.977 0.911 -0.933
95 14.76 217.86 0.888 0.622 0.7
65 -15.24 232.26 -0.917 0.707 -0.771
80 -0.24 0.06 -0.014 0 0
84 3.76 14.14 0.226 0.003 0.012
90 9.76 95.26 0.587 0.119 0.202
100 19.76 390.46 1.189 1.999 1.681
76 -4.24 17.98 -0.255 0.004 -0.017
Total 5525.8096 69.058 -18.427

COMPUTING THE KOLMOGOROV–SMIRNOV ONE-SAMPLE TEST

The Kolmogorov–Smirnov One-Sample Test is a nonparametric test used to determine whether a sample
comes from a specified distribution. It is a procedure to examine the agreement between two sets of values.
For our purposes, the two sets of values compared are an observed frequency distribution based on a randomly
collected sample and an empirical frequency distribution based on the sample’s population. Furthermore, the
observed sample is examined for normality when the empirical frequency distribution is based on a normal
distribution.

In using the test, use the steps in hypothesis testing.

STEP 1: State the Null and Research Hypotheses

The null hypothesis states that the observed sample has an approximately normal distribution. The research
hypothesis states that the observed sample does not approximately resemble a normal distribution.

STEP 2: Set the Level of Risk (or the Level of Significance) Associated with the Null Hypothesis and Choose the
Appropriate Test Statistic
The level of risk, also called an alpha (α), is frequently set at 0.05. We are seeking to compare our observed
sample against a normally distributed empirical sample. The Kolmogorov–Smirnov one-sample test will provide
this comparison.

STEP 3: Compute the Test Statistic

First, calculate the observed frequency distribution’s midpoint M and standard deviation s.
Next, use the midpoint and standard deviation to calculate the z-scores for the sample values xi,

Now, we find the relative values of the observed frequency distribution fr.

̃ and D between the cumulative frequency distributions:

Find the absolute value divergence 𝐷

Use the largest divergence with the formula below to calculate the Kolmogorov–Smirnov test statistic Z:

STEP 4: Determine and Compare the p-Value with the Level of Risk (or the Level of Significance) Associated with
the Null Hypothesis

The Kolmogorov–Smirnov test statistic Z and the Smirnov (1948) formula

If the critical value is greater than the obtained value, we must reject the null hypothesis. If the critical value is
less than the obtained p-value, we must not reject the null hypothesis.

STEP 5: Interpret the result

SAMPLE KOLMOGOROV–SMIRNOV ONE-SAMPLE TEST

A department store has decided to evaluate customer satisfaction. As part of a pilot study, the store provides customers with
a survey to rate employee friendliness. The survey uses a scale of 1–10 and its developer indicate that the scores should
conform to a normal distribution. Use the Kolmogorov–Smirnov one-sample test to decide if the sample of customers
surveyed responded with scores approximately matching a normal distribution. The survey results are shown in table below
 STEP 1: State the Null and Research Hypotheses
HO: There is no difference between the observed distribution of survey scores and a normally distributed
empirical sample.

H1: There is a difference between the observed distribution of survey scores and a normally distributed
empirical sample.

STEP 2: Set the Level of Risk (or the Level of Significance) Associated with the Null Hypothesis and Choose the
Appropriate Test Statistic
We will use an α = 0.05 in our example. In other words, there is a 95% chance that any observed
statistical difference will be real and not due to chance.
We will use the Kolmogorov–Smirnov one-sample test will provide this comparison.

STEP 3: Compute the Test Statistic

First, calculate the observed frequency distribution’s midpoint M and standard deviation s.

Now, determine the z-scores, empirical relative frequencies, and observed relative frequencies for each
score value

Score Frequency

x fi fi xi fi xi2
z-score 𝑝̂x𝑖
𝑥−𝑀
𝑧=| | (consult z-table)
𝑠
1 0 0 (0 x 0) =0 1−5.5 0.0064
| | = 2.49
1.81
2 1 2 (0 x 0) = 4 2−5.5 0.0266
| | = 1.93
1.81
3 2 6 (0 x 0) =18 3−5.5 0.0838
| | = 1.38
1.81
4 3 12 (0 x 0) =48 4−5.5 0.2033
| | = 0.83
1.81
5 6 30 (0 x 0) =150 5−5.5 0.3897
| | = 0.28
1.81
6 3 18 (0 x 0) =108 6−5.5 0.3897
| | = 0.28
1.81
7 2 14 (0 x 0) =98 7−5.5 0.2033
| | = 0.83
1.81
8 2 16 (0 x 0) =128 8−5.5 0.0838
| | = 1.38
1.81
9 1 9 (0 x 0) =81 9−5.5 0.0266
| | = 1.93
1.81
10 0 0 (0 x 0) =81 0−5.5 0.0064
| | = 2.49
1.81
To find the empirical frequency value 𝑓̂𝑟 for each value, subtract its preceding value, 𝑓̂𝑟−1 , from the associated
probability value 𝑝̂x𝑖 In other words,

𝑓̂𝑟 = 𝑝̂x𝑖 − 𝑓̂𝑟−1

We find the values of the observed frequency distribution fr

𝑓𝑖
𝑓𝑟 =
𝑛

Relative Frequency
Score Frequency Observed
Empirical Frequency
x fi Frequency
z-score 𝑝̂x𝑖 𝑓̂𝑟 = 𝑝̂x𝑖 − 𝑓̂𝑟−1 𝑓𝑖
𝑓𝑟 =
(for 1-5 then reverse) 𝑛
1 0 2.49 0.0064 0.0064 – 0.000 = 0.006 0
= 0.000
20
2 1 1.93 0.0266 0.0266– 0.006 = 0.020 1
= 0.050
20
3 2 1.38 0.0838 0.0838– 0.020 = 0.064 2
= 0.100
20
4 3 0.83 0.2033 0.2033– 0.064 = 0.140 3
= 0.150
20
5 6 0.28 0.3897 0.3897– 0.140 = 0.250 6
= 0.300
20
6 3 0.28 0.3897 0.3897– 0.140 = 0.250 3
= 0.150
20
7 2 0.83 0.2033 0.2033– 0.064 = 0.140 2
= 0.100
20
8 2 1.38 0.0838 0.0838– 0.020 = 0.064 2
= 0.100
20
9 1 1.93 0.0266 0.0266– 0.006 = 0.020 1
= 0.050
20
10 0 2.49 0.0064 0.0064 – 0.000 = 0.006 0
= 0.000
20

Next, we create cumulative frequency distributions using the empirical and observed frequency distributions. A
cumulative frequency distribution is created by taking a frequency and adding all the preceding values.

Relative Frequency Cumulative Frequency

Empirical Observed
x Empirical Frequency Observed Frequency
Frequency Frequency
𝑓̂𝑟 𝑓𝑟 𝐹̂x𝑖 𝑆x𝑖
1 0.006 0.000 0.006 0.000
2 0.020 0.050 0.020 + 0.006 = 0.026 0.050 + 0.000 = 0.050
3 0.064 0.100 0.064+ 0.026 = 0.090 0.100+ 0.000 = 0.150
4 0.140 0.150 0.140 + 0.090 = 0.230 0.150 + 0.150 = 0.300
5 0.250 0.300 0.250 + 0.230 = 0.480 0.300 + 0.300 = 0.600
6 0.250 0.150 0.250 + 0.480 = 0.730 0.150 + 0.600 = 0.750
7 0.140 0.100 0.140 + 0.730 = 0.870 0.100 + 0.750 = 0.850
8 0.064 0.100 0.064 + 0.870 = 0.934 0.100 + 0.950 = 0.950
9 0.020 0.050 0.020 + 0.934 = 0.954 0.050 + 1.000 = 1.000
10 0.006 0.000 0.006 + 0.954 = 0.960 0.000 + 1.000 = 1.000
 ̃ and D between the cumulative frequency distributions.
Now, we find the absolute value divergence 𝐷

Cumulative Frequency
Empirical Observed Divergence
x
Frequency Frequency (answers should be in absolute value)
𝐹̂x 𝑖
𝑆x𝑖 ̃ = 𝐹̂x − 𝑆x
𝐷 𝑖 𝑖 𝐷
1 0.006 0.000 0.006 - 0.000 = 0.006
2 0.026 0.050 0.026- 0.050 = 0.024 0.026 - 0.000 = 0.026
3 0.090 0.150 0.090 - 0.150 = 0.060 0.090 - 0.050 = 0.040
4 0.230 0.300 0.230 - 0.300 = 0.070 0.230 - 0.150 = 0.080
5 0.480 0.600 0.480 - 0.600 = 0.120 0.480 - 0.300 = 0.180
6 0.730 0.750 0.730 - 0.750 = 0.020 0.730 - 0.600 = 0.130
7 0.870 0.850 0.870 - 0.850 = 0.020 0.870 - 0.750 =0.120
8 0.934 0.950 0.934 - 0.950 = 0.016 0.934 - 0.850 = 0.084
9 0.954 1.000 0.954 - 1.000 = 0.046 0.954 - 0.950 = 0.004
10 0.960 1.000 0.960 - 1.000 = 0.040 0.960 - 1.000 = 0.040

̃ and D.
To find the test statistic Z, use the largest value from divergence 𝐷

̃ |, |𝐷 | ) = √20 (0.180) = 0.805

𝑧 = √𝑛 max ( |𝐷

STEP 4: Determine and Compare the p-Value with the Level of Risk (or the Level of Significance) Associated
with the Null Hypothesis

Since the critical value is less than the obtained value (0.05 < 0.536), we failed to reject the null hypothesis.

STEP 5: Interpret the result

We did not reject the null hypothesis, suggesting the customers’ survey ratings of employee friendliness sufficiently shows a
normal distribution.

Complete Table:

Relative Frequency Cumulative Frequency

Score Frequency Empirical Observed Empirical Observed
x fi fi xi fi xi2 Divergence
Frequency Frequency Frequency Frequency
z-score 𝑝̂x𝑖 𝑓̂𝑟 𝑓𝑟 𝐹̂x
𝑖
𝑆x𝑖 ̃
𝐷 𝐷
1 0 0 0 2.49 0.0064 0.006 0.000 0.006 0.000 0.006
2 1 2 4 1.93 0.0266 0.020 0.050 0.026 0.050 0.024 0.026
3 2 6 18 1.38 0.0838 0.064 0.100 0.090 0.150 0.060 0.040
4 3 12 48 0.83 0.2033 0.140 0.150 0.230 0.300 0.070 0.080
5 6 30 150 0.28 0.3897 0.250 0.300 0.480 0.600 0.120 0.180
6 3 18 108 0.28 0.3897 0.250 0.150 0.730 0.750 0.020 0.130
7 2 14 98 0.83 0.2033 0.140 0.100 0.870 0.850 0.020 0.120
8 2 16 128 1.38 0.0838 0.064 0.100 0.934 0.950 0.016 0.084
9 1 9 81 1.93 0.0266 0.020 0.050 0.954 1.000 0.046 0.004
10 0 0 0 2.49 0.0064 0.006 0.000 0.960 1.000 0.040 0.040
Assessment

Task 1. Reflection

Write a short paragraph (5–7 sentences) answering:

1. Why is testing for normality important before conducting statistical analysis?

2. How does understanding skewness and kurtosis help in interpreting data?

Score Range Description

Response is clear, well-organized, and within 5–7 sentences. Demonstrates deep understanding of
45–50 (Excellent) normality testing, skewness, and kurtosis. Concepts are accurately explained with strong insight and
clear connections to statistical analysis. Uses appropriate terminology and shows critical thinking.
Response is organized and mostly clear. Shows good understanding of concepts but may lack depth
35–44 (Very Good) or minor details. Explanations are generally accurate with slight gaps in insight or clarity. Proper
terminology is mostly used.
Response addresses the questions but lacks clarity or depth. Some concepts may be partially
25–34
explained or contain minor inaccuracies. Limited use of statistical terms. Ideas may be somewhat
(Satisfactory)
disorganized.
15–24 (Needs Response is unclear, too short/too long, or poorly organized. Demonstrates limited understanding of
Improvement) normality, skewness, and kurtosis. Contains inaccuracies or vague explanations.
0–14 (Poor) Response is incomplete, off-topic, or shows little to no understanding of the concepts.

Task 2: Test of Normality

Instructions: Consider the dataset (Math test scores) below, test the normality it. (Both methods should apply)

45, 50, 55, 60, 65, 70, 75, 80, 85, 90

Score Range Description

Accurately applies two appropriate normality tests. Complete step-by-step solution is shown with
45–50 (Excellent) correct computations and clear interpretation. Conclusions are logical and supported by results.
Demonstrates strong understanding of statistical procedures and accuracy throughout.
Applies two methods correctly with mostly accurate computations. Minor errors may be present but
35–44 (Very Good)
do not affect the overall conclusion. Interpretation is clear and mostly correct.
25–34 Attempts both methods but with noticeable errors in computation or procedure. Understanding is
(Satisfactory) evident but incomplete. Interpretation may be partially correct or unclear.
15–24 (Needs Only one method is properly attempted or both are incorrectly applied. Significant errors in
Improvement) computation and interpretation. Limited understanding of normality testing.
0–14 (Poor) Work is incomplete, incorrect, or missing. No clear attempt to apply normality tests.

References:

Demir, S. (2022). Comparison of normality tests in terms of sample sizes under different skewness and kurtosis
coefficients. International Journal of Assessment Tools in Education.

Hatem, G., Zeidan, J., Goossens, M. M., & Moreira, C. (2022). Normality testing methods and the importance of
skewness and kurtosis in statistical analysis. BAU Journal – Science and Technology.

Kim, N. (2015). Tests based on skewness and kurtosis for multivariate normality.