One – Way Classification Analysis Of Variance

Statistical comparison of at least two population means can be performed by

conducting an Analysis of Variance (ANOVA). This method partitions the total variance
of the variable of interest into several components of sources. The most basic analysis of
variance partitions the total variation into two, namely, variation due to differences in the
parent population. This type of analysis of variance is known as one-way analysis of
variance.

Assumption in the ANOVA

• The p populations from which we are sampling are normally distributed.
• The p populations have equal variances (homogeneity of variances).
• The p random samples are taken independently from one another.

Note that the data may be organized in the table below:

INDIVIDUAL
POPULATION TOTAL MEAN
1 2 … j … n1
1 𝑥11 𝑥12 … 𝑥1𝑗 … 𝑥1𝑛1 𝑥1● 𝑥̅1●
2 𝑥21 𝑥22 𝑥2𝑗 … 𝑥2𝑛2 𝑥2● 𝑥̅2●
. . . . . . .
. . . . . . .
. . . . . . .
i 𝑥𝑖1 𝑥𝑖2 ... 𝑥𝑖𝑗 … 𝑥𝑖𝑛𝑖 𝑥𝑖● 𝑥̅𝑖●
. . . . . .
. . . . . .
. . . . . .
p 𝑥𝑝1 𝑥𝑝2 … 𝑥𝑝𝑗 … 𝑥𝑝𝑛𝑝 𝑥𝑝● 𝑥̅𝑝●

Total 𝑥●●
Mean 𝑥̅●●

Procedure of Testing the p Populations:

1. 𝑯𝒐 : 𝝁𝟏 = 𝝁𝟐 = 𝝁●●● = 𝝁𝒑 , That the p population means are equal.
 𝑯𝒐 : 𝝁𝒊 ≠ 𝝁𝒌 for at least one 𝑖 ≠ 𝑘, 𝑘 = 1,2, … , 𝑝, At least two means are not
equal.
2. Test – Statistic: Use F-test at 𝛼 level of significance.
3. Decision Criterion: Reject 𝑯𝒐 𝒊𝒇 𝑭𝒄 ≥ 𝑭𝜶[(𝒑−𝟏),(𝒏−𝒑)] .
4. Computation:

Notation:
𝑥𝑖𝑗 = the ith observation in the jth group.
𝑥𝑖● = total of the observations in the ith group
𝑥●● = grand total
𝑥̅●● = grand mean

𝑥●● 𝟐
a. CF = 𝒏

𝒑 𝒏
b. TSS = ∑𝒊=𝟏 ∑𝒋=𝟏
𝒊
𝒙𝒊𝒋 𝟐 − 𝑪𝑭

𝒑 𝒙𝒊● 𝟐
c. BSS = ∑𝒊=𝟏 − 𝑪𝑭
𝒏𝒊

d. WSS = TSS – BSS

𝑩𝑺𝑺
e. MSB = (𝒑−𝟏)

𝑾𝑺𝑺
f. MSW = (𝒏−𝒑)

𝑴𝑺𝑩
g. 𝑭𝒄 =
𝑴𝑺𝑾

ANOVA Table
 Degrees of Sum Of Mean
Source of Variation
Freedom Squares Square F-Rario
(SV)
(DF) (SS) (MS)
Between Groups p–1 BSS MSB
MSB/MSW
Within Groups n–p WSS MSW
Total n–1 TSS

Example:
Three sections of the same secondary mathematics course are taught by three
teachers. The final grades were recorded as follows:
TEACHER
A B C
73 88 68
89 78 79
82 48 56
43 91 91
80 51 71
73 85 71
66 74 87
60 77 41
45 31 59
93 73 68
36 62 53
77 76 79
96 15
80
56
Is there a significant difference in the average grades given by the three teachers?
Use 0.01 level of significance.

Solution:
 1. 𝐻𝑜 : 𝜇𝐴 = 𝜇𝐵 = 𝜇𝐶 , The average grades given by the three teachers do not
differ significantly.
𝐻𝑎 : At least two of the average grades given by the three teachers differ
significantly.

2. Test-statistic: Use F-test at 1% level of significance.

3. Rejection Criterion: Reject 𝐻𝑜 if 𝐹𝑐 ≥ 𝐹0.01[(3−1),(40−3)] = 5.18.

4. Computation:
TEACHER
A B C TOTAL
73 88 68
89 78 79
82 48 56
43 91 91
80 51 71
73 85 71
66 74 87
60 77 41
45 31 59
93 73 68
36 62 53
77 76 79
96 15
80
56
n 12 15 13 40
∑𝑌 817.00 1071.00 838.00 2726.00
∑ 𝑌2 59407.00 81061.00 58994.00 199462.00
𝑌̅ 68.08 71.40 64.46 68.15

𝑥●● 𝟐 (𝟐𝟕𝟐𝟔)𝟐
a. Correction Factor = CF = = = 185776.90
𝒏 𝟒𝟎

𝒑 𝒏
b. TSS = ∑𝒊=𝟏 ∑𝒋=𝟏
𝒊
𝒚𝒊𝒋 𝟐 − 𝑪𝑭 = (732 + 892 + … + 152) – 185776.90

= 199462 – 185776.90
 = 13685.10

𝒑 𝒚𝒊● 𝟐 𝟖𝟏𝟕𝟐 𝟏𝟎𝟕𝟏𝟐 𝟖𝟑𝟖𝟐

c. BSS = ∑𝒊=𝟏 − 𝑪𝑭 = [ + + ] – 185776.90
𝒏𝒊 𝟏𝟐 𝟏𝟓 𝟏𝟑

= 186112.2526 – 185776.90
= 335.3526

d. WSS = TSS – BSS = 13685.10 – 335.3526 = 13349.7474

𝑩𝑺𝑺 𝟑𝟑𝟓.𝟑𝟓𝟐𝟔
e. MSB = (𝒑−𝟏) = = 167.6763
(𝟑−𝟏)

𝑾𝑺𝑺 𝟏𝟑𝟑𝟒𝟗.𝟕𝟒𝟕𝟒
f. MSW = (𝒏−𝒑) = = 360.8040
(𝟒𝟎−𝟑)

𝑴𝑺𝑩 𝟏𝟔𝟕.𝟔𝟕𝟔𝟑
g. 𝑭𝒄 = = 𝟑𝟔𝟎.𝟖𝟎𝟒𝟎 = 0.4647
𝑴𝑺𝑾

The resulting ANOVA table is presented below:

ANOVA
Sources of Variation Degrees of Sum of Squares Mean Square F-Ratio
(SV) Freedom (SS) (MS)
Between Groups 2 335.3526 167.6763 0.46ns
Within Groups 37 13349.7474 360.8040
Total 39 13685.1000
ns = not significant

a. Decision: Since 𝑭𝒄 = 𝟎. 𝟒𝟔𝟒𝟕 < 𝑭𝟎.𝟎𝟏[𝟐,𝟑𝟕] = 𝟓. 𝟏𝟖, there is no sufficient

evidence to reject 𝐻𝑜 .
b. Conclusion: The average grades given by the three teachers do not differ
significantly.

Two – Way Classification Analysis of Variance

 In a two – way classification, a set of observations may be classified according to
two criteria at once by means of a rectangular array in which the columns represent one
criterion of classification and rows represent a second criterion of classification. The data
are organized in a table outlined below:

INDIVIDUAL
POPULATION TOTAL MEAN
1 2 … j … c
1 𝑦11 𝑦12 … 𝑦1𝑗 … 𝑦1𝑐 𝑦1● 𝑦̅1●
2 𝑦21 𝑦22 𝑦2𝑗 … 𝑥2𝑐 𝑦2● 𝑦̅2●
. . . . . . .
. . . . . . .
. . . . . . .
i 𝑦𝑖1 𝑦𝑖2 ... 𝑦𝑖𝑗 … 𝑦𝑖𝑐 𝑦𝑖● 𝑦̅𝑖●
. . . . . .
. . . . . .
. . . . . .
r 𝑦𝑟1 𝑦𝑟2 … 𝑦𝑟𝑗 … 𝑦𝑟𝑐 𝑦𝑟● 𝑦̅𝑟●

Total 𝑦●1 𝑦●2 … 𝑦●𝑗 … 𝑦●𝑐 𝑦●●

Mean 𝑦̅●1 𝑦̅●2 … 𝑦̅●𝑗 … 𝑦̅●𝑐 𝑦̅●●

Two – Way Analysis of Variance (ANOVA) Hypothesis Testing Procedure

1. a. 𝐻𝑜 ′: 𝛼1 = 𝛼2 = … = 𝛼𝑟 , That the row means are equal.
𝐻𝑎 ′: At least two of the 𝛼1 ’s differ significantly

b. 𝐻𝑜 ": 𝛽1 = 𝛽2 = ⋯ = 𝛽𝑐 , That the column means are equal.

𝐻𝑎 ": At least two of the 𝛽1 ’s differ significantly
2. Test-Statistic: Use F-test at 𝛼 level of significance.
3. Rejection Criterion:
 a. Reject 𝐻𝑜 if 𝐹𝑐(𝑟𝑜𝑤) ≥ 𝐹𝛼[(𝑟−1),(𝑟−1)(𝑐−1)].
b. Reject 𝐻𝑜 if 𝐹𝑐(𝑐𝑜𝑙𝑢𝑚𝑛) ≥ 𝐹𝛼[(𝑐−1),(𝑟−1)(𝑐−1)].

4. Computation Formulas:
𝑦●● 𝟐
a. Correction Factor = CF = 𝒓𝒄

b. SSTotal= ∑𝒓𝒊=𝟏 ∑𝒄𝒋=𝟏 𝒚𝒊𝒋 𝟐 − 𝑪𝑭

𝒑 𝒚𝒊● 𝟐
c. SSRow = ∑𝒊=𝟏 − 𝑪𝑭
𝒄

𝒚●𝒋 𝟐
d. SSColumn = ∑𝒄𝒋=𝟏 − 𝑪𝑭
𝒄

e. SSError = SSTotal – SSRow – SSColumn

𝑺𝑺𝑹𝒐𝒘
f. MSRow = (𝒓−𝟏)

𝑺𝑺𝑪𝒐𝒍𝒖𝒎𝒏
g. MSColumn = (𝒄−𝟏)

𝑺𝑺𝑬𝒓𝒓𝒐𝒓
h. MSError = (𝒓−𝟏)(𝒄−𝟏)

𝑴𝑺𝑹𝒐𝒘
i. 𝑭𝒄(𝒓𝒐𝒘) = 𝑴𝑺𝑬𝒓𝒓𝒐𝒓

𝑴𝑺𝑪𝒐𝒍𝒖𝒎𝒏
j. 𝑭𝒄(𝒄𝒐𝒍𝒖𝒎𝒏) =
𝑴𝑺𝑬𝒓𝒓𝒐𝒓

The resulting ANOVA table is presented below:

ANOVA
 Sources of Variation Degrees of Sum of Squares Mean Square Computed
(SV) Freedom (SS) (MS) F
Row r–1 SSRow MSRow Fc(row)
Column c–1 SSColumn MSColumn Fc(column)
Error (r – 1) (c – 1) SSError MSError
Total rc – 1 SSTotal
5. State your decision based on the decision and computed F.
6. State your conclusion based on your decision.

Example for the Two – Way ANOVA

Problem: A study is made to determine the force required to pull apart pieces of glued
plastic. Three types of plastic were tested using four different levels of humidity. The
results, in newton, are given as follows:
PLASTIC HUMIDITY
TYPE 30% 50% 70% 90%
A 38.2 32.5 34.2 33.0
B 30.9 27.8 28.4 30.5
C 27.3 30.4 27.3 31.6

Use 0.05 level of significance to test the hypothesis that there is no difference in
the mean force required to pull the glued plastic apart.
a. when different types of plastic are used;
b. for different humidity condition.

Solution:
1. a. 𝐻𝑜 ′: 𝛼𝐴 = 𝛼𝐵 = 𝛼𝐶 , that there is no difference in the mean force required
to pull the glued plastic apart when different types of plastic
are used.
𝐻𝑎 ′: At least two of the plastic types differ significantly in terms of
the mean force required to pull glue plastic apart.
b. 𝐻𝑜 ": 𝛽30% = 𝛽50% = 𝛽70% = 𝛽90% , that there is no significant difference in
 the mean force required to pull the glued plastic apart among
humidity conditions.
𝐻𝑎 ": At least two of the humidity conditions differ significantly in
terms of the mean force required to pull the glued plastic apart.

2. Test-statistic: Use F-test at 𝛼 = 0.05 level of significance.

3. Decision Criterion:
a. Reject 𝐻𝑜 if 𝑭𝒄(𝒑𝒍𝒂𝒔𝒕𝒊𝒄) ≥ 𝑭𝟎.𝟎𝟓[(𝟑−𝟏),(𝟑−𝟏)(𝟒−𝟏)] = 5.14.

b. Reject 𝐻𝑜 if 𝑭𝒄(𝒉𝒖𝒎𝒊𝒅𝒊𝒕𝒚) ≥ 𝑭𝟎.𝟎𝟓[(𝟒−𝟏),(𝟑−𝟏)(𝟒−𝟏)] = 4.76.

4. Computation:
PLASTIC HUMIDITY
TOTAL MEAN
TYPE 30% 50% 70% 90%
A 38.2 32.5 34.2 33.0 137.9 34.48
B 30.9 27.8 28.4 30.5 117.6 29.40
C 27.3 30.4 27.3 31.6 116.6 29.15
TOTAL 96.4 90.7 89.9 95.1 372.1
MEAN 32.13 30.23 29.97 31.70 31.01

𝑦●● 𝟐 (372.1)𝟐
a. Correction Factor = CF = = = 11538.2008
𝒓𝒄 (𝟑)(𝟒)

b. SSTotal = ∑𝒓𝒊=𝟏 ∑𝒄𝒋=𝟏 𝒚𝒊𝒋 𝟐 − 𝑪𝑭

SSTotal = [38.22 + 32.52 + … + 31.62] - 11538.2008

= 11651.89 - 11538.2008

= 113.6892
 𝒑 𝒚𝒊● 𝟐
c. SSPlastic = ∑𝒊=𝟏 − 𝑪𝑭
𝒄
𝟏𝟑𝟕.𝟗𝟐 + 𝟏𝟏𝟕.𝟔𝟐 + 𝟏𝟏𝟔.𝟔𝟐
=[ ] - 11538.2008
𝟒
𝟒𝟔𝟒𝟒𝟏.𝟕𝟑
=[ ] - 11538.2008
𝟒

= 72.2317

𝒚●𝒋 𝟐
d. SSHumidity = ∑𝒄𝒋=𝟏 − 𝑪𝑭
𝒄
𝟗𝟔.𝟒𝟐+ 𝟗𝟎.𝟕𝟐 + 𝟖𝟗.𝟗𝟐 + 𝟗𝟓.𝟏𝟐
=[ ] - 11538.2008
𝟑
𝟑𝟒𝟔𝟒𝟓.𝟒𝟕
=[ 𝟑
]- 11538.2008

= 10.2892

e. SSError = SSTotal – SSPlastic – SSHumidity

= 113.6892 – 72.2317 – 10.2892
= 31.1683

𝑺𝑺𝑷𝒍𝒂𝒔𝒕𝒊𝒄 𝟕𝟐.𝟐𝟑𝟏𝟕
f. MSPlastic = (𝒓−𝟏)
= (𝟑−𝟏)
= 36.1159

𝑺𝑺𝑯𝒖𝒎𝒊𝒅𝒊𝒕𝒚 𝟏𝟎.𝟐𝟖𝟗𝟐
g. MSHumidity = = = 3.4297
(𝒄−𝟏) (𝟒−𝟏)

𝑺𝑺𝑬𝒓𝒓𝒐𝒓 𝟑𝟏.𝟏𝟔𝟖𝟑
h. MSError = (𝒓−𝟏)(𝒄−𝟏) = (𝟑−𝟏)(𝟒−𝟏) = 5.1947

𝑴𝑺𝑷𝒍𝒂𝒔𝒕𝒊𝒄 𝟓𝟑𝟔.𝟏𝟏𝟓𝟗
i. 𝑭𝒄(𝑷𝒍𝒂𝒔𝒕𝒊𝒄) = = = 6.9525
𝑴𝑺𝑬𝒓𝒓𝒐𝒓 𝟓.𝟏𝟗𝟒𝟕

𝑴𝑺𝑯𝒖𝒎𝒊𝒅𝒊𝒕𝒚 𝟑.𝟒𝟐𝟗𝟕
j. 𝑭𝒄(𝑯𝒖𝒎𝒊𝒅𝒊𝒕𝒚) = = = 0.6602
𝑴𝑺𝑬𝒓𝒓𝒐𝒓 𝟓.𝟏𝟗𝟒𝟕

ANOVA
 Sources of Variation Degrees of Sum of Squares Mean Square Computed
(SV) Freedom (SS) (MS) F
Plastic Type 2 72.2317 36.1159 6.95*
Humidity 3 10.2892 3.4297 0.66ns
Error 6 31.1683 5.1947
Total 11 113.6892
ns = not significant at 5% level.
* = significant at 5% level.

5. Decision Criterion:
a. Since 𝑭𝒄(𝒑𝒍𝒂𝒔𝒕𝒊𝒄) = 𝟔. 𝟗𝟓 > 𝑭𝟎.𝟎𝟓[(𝟑−𝟏),(𝟑−𝟏)(𝟒−𝟏)] = 5.14, reject 𝐻𝑜 .
b. Since 𝑭𝒄(𝒉𝒖𝒎𝒊𝒅𝒊𝒕𝒚) = 𝟎. 𝟔𝟔 < 𝑭𝟎.𝟎𝟓[(𝟒−𝟏),(𝟑−𝟏)(𝟒−𝟏)] = 4.76, accept 𝐻𝑜 .

6. Conclusion:
a. At least two of the plastic types differ significantly at 5% level in terms of the
mean force required to pull glued plastic apart.
b. The mean force required to pull the glued plastic apart do not differ at 5% level
among humidity conditions.
 ASSESSMENT TASK
Exercise
One way & Two ways ANOVA

Name: __________________________ Program & section: __________

Date: ___________________________ Score: ____________________

1. A pilot does extensive bad weather flying and decides to buy s battery-powered radio
as an independent back-up for her regular radios, which depend on the airplane’s
electrical system. She has a choice of three brands of rechargeable batteries that vary in
cost. She obtains the sample data in the following table. She randomly selects four
batteries for each brand, and test them for operating time (in hours) before recharging is
required?
BRAND OPERATING TIME (in hours)
A 20.7 21.9 20.9 22.2
B 21.0 25.6 24.7 24.5
C 26.5 26.7 25.0 24.6

2. To study the effectiveness of three kinds of packaging, a processor of a breakfast food

puts each kind into 5 different supermarkets. Use the following data representing the
number of sales of the breakfast food between 9:00 AM and noon on a given day, to test
the null hypothesis that the packaging has no effect on sales at 5% level of significance.

SUPERMARKET
PACKAGING
1 2 3 4 5
A 45 32 36 32 40
B 37 34 46 44 34
C 35 37 48 46 35