The ANOVA test is similar to the t test, except that whereas the t test compares two groups of

continuous variables with each other, the ANOVA test can compare three or more groups with one another. If
you are proficient with the t test, you are already more than halfway there to comprehending ANOVA. In terms
of SPSS procedure, the only real differences between the t test and ANOVA are in ordering the test run and
interpreting the test results; several other minor differences will be pointed out along the way.

In cases where the three pretest criteria are not satisfied for the ANOVA, the Kruskal-Wallis test, which
is conceptually similar to the ANOVA, is the better option; this alternative test is explained near the end of this
handout.

For more information about the tests, visit the following links:
https://statistics.laerd.com/spss-tutorials/one-way-anova-using-spss-statistics.php
https://statistics.laerd.com/spss-tutorials/kruskal-wallis-h-test-using-spss-statistics.php

ANALYSIS OF VARIANCE (one-way)

Example
The manufacturers of the ergonomic Able Chair want to determine the most efficient assembly
instructions for their consumers: (1) Text only, (2) Text with illustrations, or (3) Video. Which instruction
method produces the quickest consumer assembly of the Able Chair?
A researcher recruits a total of 105 participants to assemble the Able Chair. Participants
will be scheduled to come to the research center one at a time. Upon arriving, each participant will be
assigned to one of three groups (first person, Group 1; second person, Group 2; third person, Group 3; fourth
person, Group 1, and so on). Those assigned to Group 1 will be issued text-only assembly instructions, those
assigned to Group 2 will be issued assembly instructions containing text with illustrations, and those
assigned to Group 3 will be issued a video demonstrating how to assemble the chair.

1|rtqcasem2022
 Each participant will be guided to a room where there is a new Able Chair in its regular package, a
screwdriver (the only tool required), and the assembly instructions. The researcher will use a stopwatch to
time how long it takes each participant to assemble the chair.

What test is appropriate? (based on research study design)

To determine the most efficient assembly instructions for their consumers, the assembly time of the
three groups have to be compared pairwise. Further, there is only one continuous dependent variable which
is the assembly time, and only one independent variable which is the type of assembly instruction which has
three categories - Group 1 (Text only), Group 2 (Text with illustrations), and now a third group, Group 3
(Video). Thus, based on research study design, one-way ANOVA test is appropriate for the case.
The example involved timing how long it takes to put together the Able Chair using different types
of assembly instructions: Group 1 (Text only), Group 2 (Text with illustrations), and now a third group,
Group 3 (Video). The ANOVA test will reveal which (if any) of these three types of assembly instructions
statistically significantly outperforms the others in terms of how long it takes to assemble the Able Chair.

Hypotheses
The null hypothesis (H0) is phrased to anticipate that the treatment (inclusion of the illustrations or
use of video) fails to shorten the assembly time, indicating that on average, participants who are issued these
alternate forms of assembly instructions will take just as long to assemble the chair as those who are issued
the text-only instructions; in other words, there is no difference between the assembly times among these
groups. The alternative hypothesis (H 1) states that on the average, at least one group will outperform another
group:
H0: There is no difference in assembly time across the groups.
H1: At least one group outperforms another in terms of assembly time.
Admittedly, H1 is phrased fairly broadly. The post hoc Multiple Comparisons table will identify
which teaching method(s), if any, outperform which.

Data Set
Use the following data set: Example - ANOVA Test.sav. Notice that this data set has 105 records.
In this data set, records (rows) 1 through 35 are for Group 1 (Text), records 36 through 70 are for Group 2
(Text with illustrations), and records 71 through 105 are for Group 3 (Video). The data are arranged this way
just for visual clarity; the order of the records has no bearing on the statistical results.

Checking for Basic Requirements for One-Way ANOVA

The basic requirements for the ANOVA is similar to the t test except that you will assess the data for
more than two groups.
1. There is only one dependent variable – assembly time (minutes required to assemble the Able Chair),
and it is continuous. 
2. There is only one independent variable that consists of more than two groups – three independent
groups (Group 1 = Text, Group 2 = Text with illustrations, Group 3 = Video). 
3. Independence of observations – The observations in the three groups are independent because each
participant is assigned to only one of the three groups. 
4. No significant outliers
5. Normality
6. Homoscedasticity (Homogeneity of variance)
Note: The assumptions 4-6 will be assessed in each group in the data set. This is shown next.

Checking for Outliers and Normality

SPSS Procedure:
1. SPSS Syntax: Analyze > Descriptive Statistics > Explore (a dialogue box appears)
2. Transfer the dependent variable, ‘time’, into the Dependent List box and the independent
variable, ‘group’, into the Factor List box using the appropriate   buttons (or drag-and-drop the
variables into the boxes), as shown in Figure 1.
3. Click then tick Outliers (see Figure 2). (We use the default 95% confidence level
which means that the level of significance α is 0.05)
4. Click on the   button.
5. Click then tick Histogram, tick Normality plots with tests (see Figure 3).
2|rtqcasem2022
 6. Click on the   button.
7. Click on the   button. This will generate the output.

Figure 1 Figure 2

Figure 3

Checking for Outliers

Figure 4 presents the box plots of the three groups, as part of the SPSS output. There are two
outliers in the group ‘Text with illustrations’ (Group 2) – data values #38 and #50.

Figure 4 Figure 5

Guided by the methods in dealing with outliers – one of which is to replace the outlier with the
next highest/lowest value in the group’s data set. Thus, locate #38 in the data set (i.e. 110) and replace it
3|rtqcasem2022
 with the next highest data value in Group 2 (i.e. 106). Also, locate #50 (i.e. 74) and replace it with the
next lowest data value in Group 2 (i.e. 79). Do not Save the changes (as you’ll be using the same data
set later on for the Kruskal-Wallis example) or you can simply use Save As.)
Next, repeat the SPSS Procedure above. Figure 5 shows the box plots of the three groups with
the edited data set, taken from the recent SPSS output. Obviously, there are no more outliers. 

Normality Check
Table 1 presents the normality tests of the three groups, as part of the SPSS output. The Shapiro-
Wilk test revealed that the assembly times in each of the three groups follow a normal distribution as
indicated by the p values, which are all higher than 0.05 level of significance. 

Table 1

Homoscedasticity Check
Homogeneity pertains to sameness; the homogeneity of variance criterion involves checking that
the variances of the two groups are similar to each other. In SPSS, the homogeneity of variance test is
an option selected during the actual run of the ANOVA test.

SPSS Procedure:
1. SPSS Syntax: Analyze > Compare Means > One-Way ANOVA (a dialogue box appears)
2. Transfer the dependent variable, ‘time’, into the Dependent List box and the independent
variable, ‘group’, into the Factor box using the appropriate   buttons (or drag-and-drop the
variables into the boxes), as shown in Figure 6.
3. Click on the   button. This will take you to the One-Way ANOVA: Post Hoc Multiple
Comparisons window. (The post hoc test produces a table comparing the mean of each group
with the mean of every other group, along with the p value for each pair of comparisons.)
4. Tick the Tukey checkbox (see Figure 7). (As for which post hoc test to select, there are a lot of
choices. We will focus on only two options: Tukey and Sidak. Tukey is appropriate when each
group has the same n ; in this case, each group has an n of 35, so we check the ☑ Tukey
checkbox. However, if the groups had different n , then the Sidak post hoc test would be
appropriate. Further, you notice that the default level of significance α is 0.05. You can change
this depending on the set level of significance. Right now, we will use 0.05)
5. Click on the   button.
6. Click on the   button. Tick the Descriptive checkbox and the Homogeneity of variance
test, as shown in Figure 8.
7. Click on the   button. Click on the   button. This will generate the output.

Figure 6
Figure 7

4|rtqcasem2022
 If homoscedasticity is not met,
use Brown-Forsythe or Welch
test to analyze the data and use
Games-Howell test as post hoc
test.

Figure 8

The Levene’s test in the SPSS output shows that the p value is 0.069 (see Table 2). Since the
homogeneity of variance test renders a p value that is greater than .05, this tells us that there are no
statistically significant differences among the variances of the score variable for the three groups
analyzed. In other words, the variances of assembly time are similar enough among the three groups
(Text, Text with illustrations, and Video), hence we would conclude that the criterion of homogeneity of
variance has been satisfied. 

Table 2
One-Way ANOVA Test
Since all the basic requirements for ANOVA test are satisfied, then ANOVA can be used to
analyze the data set. The one-way ANOVA test was run previously when we check for
homoscedasticity requirement. So, we can obtain the one-way ANOVA test result from the same SPSS
output.

Table 3

Now, we look at the ANOVA table (Table 3) and find a Sig. ( p) value of .001; since this is less
than the α level of .05, this tells us that there is a statistically significant difference among the (three)
group means of assembly times, but unlike reading the results of the t test, we are not done yet.

Remember that in the realm of the t test, there are only two groups involved, so interpreting the
p value is fairly straightforward: If p ≤ .05, there is no question as to which group is different from
which—clearly, the mean of Group 1 is statistically significantly different from the mean of Group 2.
However, when there are three or more groups, we need more information to determine which group is
different from which; that is what the post hoc test answers.

The ANOVA table (Table 3) produced a statistically significant p value (Sig. = .001), which
indicates that there is a statistically significant difference detected somewhere among the three groups;
the post hoc table will tell us precisely which pairs are statistically significantly different from each

5|rtqcasem2022
 other. Specifically, it will reveal which group(s) outperformed which. This brings us to the (Tukey post
hoc) Multiple Comparisons table (Table 4).

Means are shown in Table 5 for Group 1 (Text), Group 2 (Text with illustrations), and Group 3
(Video). We will use Table 5 (Descriptives) and Table 4 (Multiple Comparisons) to analyze the
ANOVA test results. Table 5 lists the mean time for each of the three groups: μ(Text) = 97.63, μ(Text
with illustrations) = 92.03, and μ(Video) = 89.77. We will assess each of the three pairwise score
comparisons separately.

Table 4

Table 5

Comparison 1— Text versus Text With Illustrations

Table 4 first compares the mean time for the Text group with the mean score for the Text with
illustrations group, which produces a Sig. ( p) of .020. Since p is less than the .05 α level, this tells us
that for assembly time, there is a statistically significant difference between Text(μ = 97.63) and Text
with illustrations (μ = 92.03). As it generated lesser assembly time, this implies that text with
illustrations is more effective instruction method compared with text only.

Comparison 2— Text versus Video

The second comparison in Table 4 is between Text and Video, which produces a Sig. ( p)
of .001. Since p is less than the .05 α level, this tells us that for assembly time, there is a statistically
significant difference between Text (μ = 97.63) and Video (μ = 89.77). This implies that video
instruction is more effective method compared with text only.

Comparison 3— Text With Illustrations versus Video

The third comparison in Table 4 is between Text with illustrations and Video, which produces a
Sig. ( p) of .515. Since p is greater than the .05 α level, this tells us that for assembly time, there is no
statistically significant difference between Text with illustrations (μ = 92.03) and Video (μ = 89.77).
Since there is no significant difference in the assembly time of the two groups, this implies that text with
illustrations is as effective as video instruction.

This concludes the analysis of the Multiple Comparisons (post hoc) table. You have probably
noticed that we skipped analyzing half of the rows; this is because there is a double redundancy among
the figures in the Sig. column. This is the kind of double redundancy you would expect to see in a
6|rtqcasem2022
 typical two-dimensional table. The Sig. column of the Multiple Comparisons table (Table 4) contains
two p values of .020: one comparing Text to Text with illustrations and the other comparing Text with
illustrations to Text. In addition, there are two p values of .001 (Text vs Video and Video vs Text) and
two p values of .515 (Text with illustrations vs Video and Video vs Text with illustrations).

Effect Size for One-Way ANOVA (partial eta squared η2 )

To generate the effect size called partial eta squared (η2 ) for a one-way ANOVA, you need to
run the GLM procedure, as shown below:
SPSS Procedure:
1. SPSS Syntax: Analyze > General Linear Model > Univariate (a dialogue box appears)
2. Transfer the dependent variable, ‘time’, into the Dependent Variable box and the independent
variable, ‘group’, into the Fixed Factor(s) box using the appropriate   buttons, as shown in
Figure 9.
3. Click the   button. You will be presented with the Univariate: Options dialogue box.
4. Select the Estimates of effect size checkbox, as shown in Figure 10. (The level of significance is
the default value, .05)
5. Click on the   button.
6. Click on the   button. This will generate the output.

Figure 9

Figure 10

Table 6 is a portion of the generated SPSS output in the GLM procedure. The effect size which
is the partial eta squared value is η2 = .133.

Table 6

7|rtqcasem2022
 Documenting Results
“A one-way ANOVA was conducted to determine if the assembly time was
different for groups with different types on instruction. Participants were classified into
three groups: text only (n = 35), text with illustrations (n = 35), and video (n = 35). There
were no outliers in the data, as assessed by inspection of the boxplots; the assembly time
was normally distributed in each of the three groups according to the type of instruction, as
assessed by Shapiro-Wilk's test ( p > .05); and there was homogeneity of variances, as
assessed by Levene's test of homogeneity of variances ( p = .069).
Data is presented as mean ± standard deviation. There was a statistically significant
difference on the assembly time between groups as indicated by one-way ANOVA
(F(2,102) = 7.795, p = .001, η2 = .133). A Tukey post hoc test revealed that the time to
assemble the ergonomic Able Chair was statistically significantly lower using the text with
illustrations (92.03 ± 7.318 min, p = .020) and video instruction (89.77 ± 8.296 min, p
= .001) compared to text only (97.63 ± 9.903 min). There was no statistically significant
difference between the ‘text with illustrations’ and ‘video’ instruction ( p = .515).”
Note: the numerical values in the documentation result were extracted from Tables 2-6.

All files including data sets can be downloaded from class drive:
https://drive.google.com/drive/folders/1L7L4K5frFd0M2RNHSeoXWntnAHdOpIqV?usp=sharing
For SPSS video tutorials on ANOVA test, visit the following link:
https://www.statstutor.ac.uk/resources/uploaded/statstutor--one-way-anova-in-spss-v2.mp4

8|rtqcasem2022