Name: LEARNING ACTIVITY SHEET 4

One-Way Analysis of Variance

Analysis of Variance (ANOVA)

Supposed we are interested in the effect of three teaching approaches (conventional, modular, blended
learning) to the performance of the students in Mathematics. If there is no difference between the different
approaches, then we would expect all of them to be approximately equal. Otherwise, we would expect
the mean performances to differ. Here, we introduce a situation that utilizes a test that is used to analyze
the data from more than two populations. These tests are used to deal with treatment (e.g. teaching
approaches) effects, including tests that take into account other factors that may affect the response (e.g.
Mathematics performance). The hypothesis that the population means are equal is considered equivalent
to the hypothesis that there is no difference in treatment effects. The analytical method we will use in such
problems is called the analysis of variance (ANOVA). Initial development of this method could be credited
to Sir Ronald A. Fisher who introduced this technique for the analysis of agricultural field experiments which
at present is also useful when dealing with social science and educational researches.

AS_4a.
Discuss briefly the history and works of Sir Ronald A. Fisher.

One-way ANOVA

We are going to study the hypothesis testing problem of comparing population means for more than
two independent populations, where the data are about several independent groups (different
treatments being applied, or different populations being sampled).
In one-way ANOVA, it is assumed that the 𝑘 populations are independent and normally distributed with
means 𝜇1 , … , 𝜇𝑘 and with unknown but equal variance 𝜎 2 . The question is whether the means of these
groups are different or equal. That is, we test the hypothesis that:

Ho: 𝜇1 = 𝜇2 = ⋯ = 𝜇𝑘 versus
Ha: At least two of the means are different.

The idea is to consider the overall variability in the data and partition it into two parts: (1) between-
groups variability and (2) within-groups variability. If the between groups is much larger than that within
groups, this will indicate that differences between the groups are real. Let independent samples be
drawn randomly of sizes 𝑛1 , 𝑛2 , … 𝑛𝑘 and let 𝑛𝑡 = 𝑛1 + 𝑛2 + ⋯ + 𝑛𝑘 .

Numerous Statistics books discuss the mathematics behind this statistical tool but this module will spare
you from analyzing those and we get straightly to the point on how to calculate the F-statistic (maybe
attributed from Fisher hence also called the F-value) manually. First consider the following ANOVA table,

Advanced Statistics /LAS 4/Page 1 of 6

 Source of Degrees of
Sum of Squares Mean Squares F-statistic
Variation Freedom
(∑ 𝑥)2 (∑ ∑ 𝑥)2 𝑆𝑆𝑏 𝑀𝑆𝑏
Between 𝑆𝑆𝑏 = ∑ − 𝑘−1 𝑀𝑆𝑏 = 𝐹=
𝑛 𝑛𝑡 𝑘−1 𝑀𝑆𝑤
(∑ 𝑥)2 𝑆𝑆𝑤
Within 𝑆𝑆𝑤 = ∑ ∑(𝑥 2 ) − ∑ 𝑛𝑡 − 𝑘 𝑀𝑆𝑤 =
𝑛 𝑛𝑡 − 𝑘
Total 𝑆𝑆𝑡 = 𝑆𝑆𝑏 + 𝑆𝑆𝑤 𝑛𝑡 − 1

Notice that the table must be filled out from left to right based on the flow of the formulae. It may seem
overwhelming also to perform the calculations but to facilitate understanding of the process of
calculation, we illustrate it using an example.

Example 2.2.1. The samples presented in the table below (blue-colored part) represent test scores
from three classes of Statistics taught using three different approaches (conventional,
modular and blended learning) and are independently obtained. Assume that the three
different populations are normal with equal variances. At α = 0.05 level of significance, test
for equality of population means.

In this problem, we will test the hypothesis that:

𝐻𝑜: The mean test scores of the class are equal.
𝐻𝑎: At least two of the classes in Statistics differ in test scores.
Class 1 Class 2 Class 3
2 2
𝑥1 𝑥1 𝑥2 𝑥2 𝑥3 𝑥3 2
64 4096 56 3136 81 6561
84 7056 74 5476 92 8464
75 5625 69 4761 84 7056
77 5929
80 6400 𝑘 = 3 groups

𝑛 5 3 3 𝑛𝑡 = ∑ 𝑛 = 11

∑𝑥 380 199 257 ∑ ∑ 𝑥 = 836

(∑ ∑ 𝑥)2 8362
𝑥̅ 76 66.33 85.67 𝑛𝑡
= 11
= 63536

(∑ 𝑥)2 (∑ 𝑥)2
28880 13200.33 22016.33 ∑ = 64096.67
𝑛 𝑛

∑(𝑥 2 ) 29106 13373 22081 ∑ ∑(𝑥 2 ) = 64560

Step 1. Square each of the given values to obtain 𝑥𝑘 2 (e.g. 𝑥1 2 , 𝑥2 2 , 𝑥3 2 )

Step 2. Perform summations (orange cells).
Step 3. Summate the 𝑛’s (to obtain 𝑛𝑡 ) and ∑ 𝑥’s (to obtain ∑ ∑ 𝑥).
(∑ ∑ 𝑥)2
Step 4. Square ∑ ∑ 𝑥 from step 3 and divide by 𝑛𝑡 to obtain . 𝑛𝑡
(∑ 𝑥)2 (∑ 𝑥)2
Step 5. Summate the ’s (to obtain ∑ ) and ∑(𝑥 2 )’s (to obtain ∑ ∑(𝑥 2 ))
𝑛 𝑛

Notice that the red-colored cells contain the values needed in calculating the “Sum of Squares” in
our ANOVA table. Therefore, the ANOVA table will be,
Source of Degrees of
Sum of Squares Mean Squares F-statistic
Variation Freedom
𝑆𝑆𝑏 𝑀𝑆𝑏
(∑ 𝑥)2 (∑ ∑ 𝑥)2 𝑀𝑆𝑏 = 𝐹=
𝑆𝑆𝑏 = ∑ − 𝑘−1 𝑘−1 𝑀𝑆𝑤
𝑛 𝑛𝑡
Between = 64096.67 − 63536 = 3−1 560.67 280.335
= =
= 560.67 =2 2 57.916
= 280.335 = 4.840
𝑆𝑆𝑤
(∑ 𝑥)2 𝑀𝑆𝑤 =
𝑆𝑆𝑤 = ∑ ∑(𝑥 2 ) − ∑ 𝑛 𝑛𝑡 − 𝑘 𝑛𝑡 − 𝑘
Within = 64560 − 64096.67 = 11 − 3 463.33
=8 =
= 463.33 8
= 57.916
𝑆𝑆𝑡 = 𝑆𝑆𝑏 + 𝑆𝑆𝑤 𝑛𝑡 − 1
Total = 560.67 + 463.33 = 11 − 1
= 1024 = 10
Advanced Statistics /LAS 4/Page 2 of 6
 Now, we need to decide whether to accept or reject the null hypothesis. Therefore, we need to
complete our table,
Source of Sum of Degrees of Mean F- F- P-
Variation Squares Freedom Squares statistic critical value
Between 560.67 2 280.335 4.840 4.459 0.042
Within 463.33 8 57.916
Total 1024 10

The F-critical can be obtained by looking up on the table of critical values for an F-distribution. However,
the Microsoft Excel can help you obtain such by entering the formula:
= F. INV. RT (α, dfb , dfw ) = F. INV. RT (0.05,2, 8)
F-critical
It will return 4.840
Note that when using this method in deciding whether to accept or reject the hypothesis, if:
• F-statistic < F-critical, accept the Ho.
• F-statistic > F-critical, reject the Ho and conclude the Ha.
Since 4.840 > 4.459, we reject the null hypothesis and conclude that at least 2 of the classes in
Statistics differ in test scores. Similarly,

= FDIST(F − statistic, dfb , dfw ) = FDIST(4.840,2,8)

P-value
It will return 0.042
Note that when using this method in deciding whether to accept or reject the hypothesis, if:
• P-value < α, reject the Ho and conclude the Ha.
• P-value > α, accept the Ho.
Since we set 𝛼 = 0.05 and 0.042 < 0.05, we reject the null hypothesis and conclude that at least 2 of
the classes in Statistics differ in test scores.

Either way, we provide the same conclusion as long as the interpretation is correct. Note also that in
using ANOVA as a statistical tool, you must first satisfy the following assumptions attributed to it,
otherwise, we must use non-parametric test which will be discussed on the succeeding units of this
subject.
1. Normality – that each sample is taken from a normally distributed population
2. Sample Independence – that each sample has been drawn independently of the other samples
3. Variance Equality – that the variance of data in the different groups should be the same
4. Continuous Dependent Variable – that is measured on a scale which can be subdivided using
increments/ intervals

AS_4b.
In an experiment to investigate the long-distance marathon performance of the four (4) colleges, 5 students
from each college were taken and the combined distances in kilometers ran by each group at a given
time was recorded. A partially completed ANOVA table is given. Fill in the missing entries (blue cells) and
test the relevant hypothesis using 0.05 level of significance

Source of Sum of Degrees of Mean F- F- P-

Variation Squares Freedom Squares statistic critical value
Between
Within 𝟐𝟑𝟓𝟒𝟏𝟗. 𝟎𝟒

Total 𝟑𝟏𝟎𝟓𝟎𝟎. 𝟕𝟔

1. Formulate the null and alternative hypothesis.

𝐻𝑜 :
𝐻𝑎 :
2. Decide the level of significance, 𝒂.
The hypothesis will be tested using 𝑎 = ______ level of significance.

Advanced Statistics /LAS 4/Page 3 of 6

3. Choose the appropriate test statistic.
The test that will be used is ___________________________________ because __________________________
_______________________________________________________________________________________________.
4. Compute the value of the statistical test.
Computation:
(Show calculation for each of the missing entries.)
Source of Sum of Degrees of
Mean Squares F-statistic F-critical P-value
Variation Squares Freedom

Between

Within 𝟐𝟑𝟓𝟒𝟏𝟗. 𝟎𝟒

Total 𝟑𝟏𝟎𝟓𝟎𝟎. 𝟕𝟔

5. Decide whether to accept or reject the null hypothesis.

Decision: Reject Ho Accept Ho Decision: Reject Ho Accept Ho

Reason: Reason:

6. Draw a conclusion.
Conclusion:

Advanced Statistics /LAS 4/Page 4 of 6

 AS_4c.
A researcher quantified the perceptions of Junior High School students towards the effect of technology in
learning using 5-point Likert scale. He involved 46 student-respondents of which 11 are Grade 7; 9 are Grade
8; 12 are Grade 9; and, 14 are Grade 10. The result of data collection is posted below. At 0.05 level of
significance, is there a significant difference between the students’ perceptions when they are grouped
according to their grade level? Follow the steps in hypothesis testing procedure.
Grade 7 Grade 8 Grade 9 Grade 10
4.67 3.80 3.67 2.80
3.27 3.73 2.93 2.67
3.27 3.73 3.67 2.87
2.87 3.33 3.27 3.07
2.60 3.13 3.20 2.87
3.93 4.27 3.20 3.53
3.47 3.93 3.33 2.47
3.80 3.07 3.07 3.33
3.40 3.60 2.87 3.07
3.07 3.27 4.07
3.00 2.80 3.67
3.40 3.27
3.60
3.47

Advanced Statistics /LAS 4/Page 5 of 6

Advanced Statistics /LAS 4/Page 6 of 6