MODULE 3

ANALYSIS OF VARIANCE
Definition
Analysis of variance (ANOVA) is a statistical technique specially designed to test
whether the means of more than two quantitative populations are equal.
Types of ANOVA
One-way ANOVA
It is a statistical method used to determine whether there are statistically
significant differences between the means of three or more independent groups.
It is often used when testing the effects of a single independent variable.
ex: Fertilizer is used as an independent variable used to find out the crop yield of
different mixes.
Two-way ANOVA
It is a statistical test used to determine how two independent variables influence
a continuous dependent variable, both individually and in combination.
ex: Fertilizer and type of seed is used is used to find out the crop yield of different
mixes.
Problem
[Link] is the data of 5 variety of plots and 4 crops. Analyse the variance whether
there is significant difference in yield of crops.
P1 P2 P3 P4 P5
C1 42 50 62 34 52
C2 48 66 68 78 70
C3 68 52 76 64 70
C4 80 94 78 82 66
Solution:
C1 C2 C3 C4
P1 42 48 68 80
P2 50 66 52 94
P3 62 68 76 78
P4 34 78 64 82
P5 52 70 70 66
240 330 330 400
Step 1: Setting Ho
Ho- there is no significant difference in mean yield of different crops.
H1- there is significant difference in mean yield of different crops.
Step 2: Test criteria
Since we are comparing the mean between more than 2 groups, we use one way
anova.
Step 3: Test statistics
T = sum of observations
T = 1300
T2/n = 13002/20 (n = R x C)
= 84500
Sum of squares of totals (SST) = sum of square of all observations – T2/n
C1 C2 C3 C4
P1 422 482 682 802
P2 502 662 522 942
P3 622 682 762 782
P4 342 782 642 822
P5 522 702 702 662

C1 C2 C3 C4
P1 1764 2304 4624 6400
P2 2500 4356 2704 8836
P3 3844 4624 5776 6084
P4 1156 6084 4096 6724
P5 2704 4900 4900 4356

Sum of squares of totals (SST) = Sum of square of all observations – T2/n

= 88736 – 84500
= 4236
Sum of square of columns (SSC) = ∑x12/n1 + ∑x22/n2 + ∑x32/n3 +…+ ∑xn2/nn
= 2402/5 + 3302/5 + 3302/ 5 + 4002/5 – 84500
= 11520 + 21780 + 21780 + 32000 – 84500
= 2580
SSE (sum of squared estimation of error) = SST – SSC
= 4236 – 2580
= 1656
Anova table
Variance Sum square DF Mean F value
Between SSC K-1 MSC= SSC/K-1 MSC/MSE
sample OR

Within SSE N-K MSE = SSE/N-K MSE/MSC

sample

About sample SST N-1

variance Sum square DF Mean F value

Between 2580 4 – 1= 3 860 860/ 103.5
sample
Within 1656 20 – 4 = 16 103.5 = 8.309
sample

About sample 4236 20 – 1 = 19

NOTE:
It is to be understand that whichever is the higher value becomes the numerator.
here f value = MSC/MSE
Step 4: Level of significance and DOF
LOS: 5%
DOF = K -1, n-k = 3,16
Step 5: Table value
From f table, Critical value = 3.239
Step 6: Comparison of values
Ts>Tv (8.309>3.239), we reject Ho there is significant difference in mean yield
of different crops
Coding method in ANOVA
[Link] whether there is a significant difference between the mean yield of four
varieties of wheat on 5 plots.
Plot/varieties A B C D
1 99 103 104 104
2 101 102 103 100
3 103 100 107 103
4 99 105 97 107
5 98 95 99 106

Plot/varieties A B C D
1 -1 3 9 4
2 1 2 3 0
3 3 0 7 3
4 -1 5 -3 7
5 -2 -5 -1 6
0 5 15 20

Step 1: Setting Ho
Ho: sample means are equal
Step 2: Test criteria
Since we are comparing the mean between more than 2 groups, we use one way
anova
Step 3: Test statistics
T = sum of observations
= 0 + 5 + 15 + 20
= 40
N = 5 x 4 = 20
T2/N = 402/20 = 80
Sum of squares of totals (SST) = sum of square of all observations – T2/n
 Plot/varieties A B C D
1 1 9 81 16
2 1 4 9 0
3 9 0 49 9
4 1 25 9 49
5 4 25 1 36
16 63 149 110
Sum of squares of totals (SST) = Sum of square of all observations – T2/n
= 338 – 80
= 258
Sum of square of columns (SSC) = ∑x12/n1 + ∑x22/n2 + ∑x32/n3 +…+ ∑xn2/nn
= 02/5 + 52/5 + 152/ 5 + 202/5 – 80
= 0+ 5 + 45 + 80 – 80
=50
SSE (sum of squared estimation of error) = SST – SSC
= 258 – 50
= 208
Anova table
Variance Sum square DF Mean F value
Between SSC K-1 MSC= SSC/K-1 MSC/MSE
sample OR

Within SSE N-K MSE = SSE/N-K MSE/MSC

sample

About sample SST N-1

Variance Sum square DF Mean F value

Between 50 4 – 1= 3 16.67 16.67/ 13
sample
Within 208 20 – 4 = 16 13 = 1.28
sample

About sample 258 20 – 1 = 19

NOTE:
It is to be understand that whichever is the higher value becomes the numerator.
here f value = MSC/MSE
Step 4: Level of significance and DOF
LOS: 5%
DOF = (K -1, n - k) = 3,16
Step 5: Table value
From f table, Critical value (Tv) = 3.239
Step 6: Comparison of values
Ts<Tv (1.28 < 3.239), we accept Ho therefore the sample means are equal.
Problem 3
The following data is regarding 3 varieties of wheat grown on 4 plots
A B C

P1 6 5 5

P2 7 5 4

P3 3 3 3

P4 8 7 4

A,B,C are varieties of crops and P1,P2,P3,P4 are different [Link] whether
there is significant difference between yield varieties of wheat and difference in
plot.
Solution
Step 1
Setting Hypothesis
Null Hypothesis(H0)
H0C: There is no significant difference between mean yield of variety of wheat.
or
Column means are equal.
H0R: There is no significant difference between mean yield in different plots.
or
Row means are equal.
Alternate Hypothesis(H1)
H1C: There is a significant difference between mean yield of variety of wheat.
or
Column means are not equal.
H1R: There is a significant difference between mean yield in different plots.
or
Row means are not equal.
Step 2

A B(XC2) C(XC3) ∑XR

(XC1)
P1(XR1) 6 5 5 16
P2(XR2) 7 5 4 16
P3(XR3) 3 3 3 9
P4(XR4) 8 7 4 19
∑XC 24 20 16 -----
-

N([Link] observations) = 12 C=3 (No. of columns) and R=4 (No. of rows)

N=3x4
Sum of all observations(T) = 6+5+5+7+5+4+3+3+3+8+7+4 = 60
Correction Term(CT) = T2/N = 602/12 = 300
Step 3
Sum of squares of total (SST) = Sum of squares of observations - CT
Sum of squares of observations = (62+52+52+72+52+42+32+32+32+82+72+42)-300
= 32
Sum of squares of columns (SSC)
= ((∑XC1)2/nc1)+( ∑XC2)2/nC2)+( ∑XC3)2/nC3))-CT
2
= ((24 /4)+(202/4)+(162/4))-300 =8

Sum of squares of rows (SSR)

= ((∑XR1)2/nR1)+( ∑XR2)2/nR2)+( ∑XR3)2/nR3)+ (∑XR4)2/nR4))-CT
= ((162/3)+(162/3)+(92/3)+(192/3))-300 = 18

Sum of squares within sample (SSE) = SST-SSC-SSR

= 32-8-16= 6
Step 4
Mean squares due to columns (MSC) = (SSC)/(C-1)
=(8)/(3-1)= 4
Mean squares due to rows (MSR) = (SSR)/(R-1)
= (18)/(4-1) = 6
Mean squares due to sample (MSE) = (SSE)/(C-1)(R-1)
= (6)/(3-1)(4-1)= 1
Step 5
Find the value of F due to column and row
FC = MSC/MSE = 4/1= 4
FR = MSR/MSE = 6/1 = 6

STEP 6
ANOVA TABLE

Source of Sum of Degree of Mean square F

variance square freedom
Between SSC = 8 C-1 = 3-1 = 2 MSC= 4 FC=4
columns
Between rows SSR = 18 R-1 = 4-1= 3 MSR= 6 FR=6

Within sample SSE = 6 (C-1)(R-1) = MSE = 1

(3-1)(4-1)= 6
Total SST = 32
Step 7
Find the value of F from table
Here take level of significance α = 5%
F value due to column F value due to row
Degree of freedom df, Degree of freedom df,
df = ((C-1),(C-1)(R-1)) df = ((R-1),(R-1)(C-1))
=((3-1),(3-1)(4-1)) =((4-1),(4-1)(3-1)
df(2,6) df(3,6)
Ftable = 5.14 Ftable = 4.76
Here calculated value is less than table Calculated value is greater than
value table value
4<5.14 6>4.76
Therefore null hypothesis(H0) is Therefore null hypothesis(H0)
Accepted is Rejected, Alternate hypothes
is(H1) is Accepted

Problem 4
In a factory, production is carried on by 5 workers, each using 4 types of
machines A,B,C,[Link] following table gives number of units produced perday
by the five workers using the 4 machines.
Workers A B C D

1 46 40 49 38

2 48 42 54 45

3 36 38 46 34

4 46 40 48 35
 5 40 45 51 41

Test whether mean productivity of machines are same.

Test whether the 5 workers differ significantly with respect to mean
productivity.
Solution
Here observations are large numbers, for convenience the numbers are reduced
by using coding method. In these problem observations are subtracted from 41.
Workers A B C D

1 5 -1 8 -3

2 7 1 13 4

3 -5 -3 5 -7

4 5 -1 7 -6

5 -1 4 10 0

Step 1
Setting Hypothesis
Null Hypothesis(H0)
H0C: Column means are equal
H0R: Row means are equal
Alternate Hypothesis(H1)
H1C: Column means are not equal
H1R: Row means are not equal
Step 2

Workers A(XC1) B(XC2) C(XC3) D(XC4) ∑XR

1 (XR1) 5 -1 8 -3 9
2(XR2) 7 1 13 4 25

3(XR3) -5 -3 5 -7 -10

4(XR4) 5 -1 7 -6 5
5(XR5) -1 4 10 0 13
∑XC 11 0 43 -12 ---------
N=20 C=4 R=5
Sum of all observations(T) = 5+7-5+5-1-1+1-3-1+4+8+13+5+7+10-3+4-7-
6+0= 42
Correction Term(CT) = T2/N = 422/20= 88.2
Step 3
Sum of squares of total (SST) = Sum of squares of observations - CT
Sum of squares of observations = (52+72-52+52-12-12+12-32-
12+42+82+132+52+72+102-32+42-72-62+02)-88.2= 581.8
Sum of squares of columns (SSC)
= ((∑XC1)2/nc1)+( ∑XC2)2/nC2)+( ∑XC3)2/nC3)+( ∑XC4)2/nC4))-CT
= ((112/5)+(02/5)+(432/5)+(-122/5)-88.2=334.6
Sum of squares of rows (SSR)
= ((∑XR1)2/nR1)+( ∑XR2)2/nR2)+( ∑XR3)2/nR3)+ (∑XR4)2/nR4)+ (∑XR5)2/nR5) )-
CT
2
= ((9 /4)+(252/4)+(-102/4)+(52/4)+ (132/4))-88.2 = 161.8
Sum of squares within sample (SSE) = SST-SSC-SSR
= 581.8-334.6-161.8 = 85.4
Step 4
Mean squares due to columns (MSC) = (SSC)/(C-1)
 =(334.6)/(4-1)= 111.53
Mean squares due to rows (MSR) = (SSR)/(R-1)
= (161.8)/(5-1) = 40.45
Mean squares due to sample (MSE) = (SSE)/(C-1)(R-1)
= (85.4)/(4-1)(5-1)= 7.11
Step 5
Find the value of F due to column and row
FC = MSC/MSE = 111.53/7.11= 15.68
FR = MSR/MSE = 40.45/7.11 = 5.68
STEP 6
ANOVA TABLE

Source of Sum of Degree of Mean square F

variance square freedom
Between SSC = 334.6 C-1 = 4-1 = 3 MSC= 111.53 FC=15.68
columns
Between rows SSR = 161.8 R-1 = 5-1= 4 MSR= 40.45 FR=7.11

Within sample SSE = 85.4 (C-1)(R-1) = MSE = 7.11

(4-1)(5-1)= 12
Total SST = 581.8

Step 7
Find the value of F from table
Here take level of significance α = 5%
F value due to column F value due to row
Degree of freedom df, Degree of freedom df,
df = ((C-1),(C-1)(R-1)) df = ((R-1),(R-1)(C-1))
=((4-1),(4-1)(5-1)) =((5-1),(5-1)(4-1))
df(3,12) df(4,12)
Ftable = 3.49 Ftable = 3.26
Calculated value is greater than Calculated value is greater than
table value table value
15.68>3.49 7.11>3.26
Therefore null hypothesis(H0) Therefore null hypothesis(H0)
is Rejected, Alternate hypothes is Rejected, Alternate hypothes
is(H1) is Accepted is(H1) is Accepted
 RANDOMISED COMPLETE BLOCK DESIGN

Analysis of Variance
Suppose in investigating the performance of the 3 types of cars, 5 drivers were each
assigned one car of each type. In statistics, this is called a randomised complete
block design, consisting of b=5 blocks, with k=3 treatments. Each block contains
matched groups of k experimental units – one unit for each treatment - so in this
example there are three experimental units in each block.

The analysis of variance for an RCBD partitions the total sum of squares into three
parts:

The ANOVA table:

SOURCE df SS MS F p
Blocks
Treatment
Error
Total

Note that the table is similar to the Completely Randomised ANOVA table, but
there is now a row included for blocks. b is generally used to represent the
number of blocks involved. Using this added information, the table can be
completed:

Assumptions of the RCBD:

1) Sampling: a. The blocks are independently sampled

b. The treatments are randomly assigned to the experimental units within block.

2) Homogeneous Variance: The treatments all have the same variability,

i.e. they all have the same variance

3) Approximate Normality: Each population is normally distributed

ANOVA CALCULATIONS FOR RANDOMISED BLOCK DESIGN

The calculations are similar, but note the differences.

• Calculate the Degrees of Freedom for each source of variation:

dfblocks = (b − 1); dftreatment = (k − 1); dftotal = (n − 1);

df error = df total − df treatment − df blocks = (n − k − b + 1)

• Calculate the Correction Factor

• Calculate the Total Sum of Squares

• Calculate the Block totals

• Calculate the Treatment totals

• Calculate the Block Sum of Squares

• Calculate the Group (Treatment) Sum of Squares

• Calculate the Error Sum of Squares (by subtraction)

• Calculate the Mean Square for Block, Group and Error – divide

SSblock , SStreatmenet and SS error by b − 1, k − 1 and (n − b − k + 1)

• Calculate the F Value

• Determine the p-value

 ANOVA Table for randomised complete block design
Correction Factor: ( all entries ) 2
( y) 2 ▪ Add all entries
CF = = ▪ Square the result
number of entries n ▪ Divide by n
Total Sum of ▪ Square each entry
Squares: (y ) 2
▪ Add all results
= (each entry) − CF = y −
2 2
SS ▪ Subtract CF
Total
n
Sum of Squares • Square each
between Blocks: ∑(𝐵𝑙𝑜𝑐𝑘 𝑡𝑜𝑡𝑎𝑙)2 block total
SSBlocks = 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 − 𝐶𝐹 • Add totals
• Divide by no. of
treatments
𝐵 21+𝐵22+⋯+𝐵2𝑘 (∑ 𝑦)2 • Add results
= − • Subtract CF
𝑘 𝑛
This Sum of Squares
reflects the
differences between
the blocks
Sum of Squares ∑(𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 𝑡𝑜𝑡𝑎𝑙)2
between SSTreatments= − 𝐶𝐹 ▪ Square each
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑙𝑜𝑐𝑘𝑠 group (Treatment)
Treatments
total
𝑇 12+𝑇22+⋯+𝑇2𝑘 (∑ 𝑦)2 ▪ Divide by no. of
= − blocks
𝑏 𝑛 ▪ Add results
▪ Subtract CF

Sum of Squares for

Error: SS Error
= SS Total
− (SS Blocks
+ SS Treatmentst
)
Mean Square for SS Blo cks
Blocks: MSBlocks =
b −1
Mean Square for SS
Treatments: MSTreatments = Trea tm en ts
k −1
Mean Square for SS Error
MS Error =
(n − b − k + 1)
Error:

Test statistic F: MSTreatm ents MS

FTreatments = F =
Blo cks

Block
MS Error MS Error

Standard Error of 2 × 𝑀𝑆𝐸𝑟𝑟𝑜𝑟 b is number of blocks

Difference between 𝑆𝐸𝐷(𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠) = √
𝑏 r is number of values
means:
used to calculate the
Note General Formula: 𝑆𝐸𝐷 = √2 × 𝑀𝑆𝐸𝑟𝑟𝑜𝑟 respective mean.
𝑟
Returning to the Car Brand Problem:
Driver Total
M N P R S
Brand A 7.6 8.4 8.0 7.6 8.4 40.0
Brand B 7.8 8.0 9.1 8.5 9.6 43
Brand C 9.6 10.4 9.2 9.7 10.6 49.5
Totals 25 26.8 26.3 25.8 28.6 132.5

(i) df block = 4; df treatment = 2; df total = 14; df error = df total − df block − df treatment = 8

(ii) Calculate the correction factor

CF = ( all entries ) 2

=
( y) 2
= 1170.42 .

number of entries n
(iii) Calculate the Total Sum of Squares ( y ) 2

SSTotal =  (each entry ) − CF =  y − = 13.69

2 2

n
(iv) Calculate the Block Sum of Squares
(25 2
+ 26.82 + 26.32 + 25.82 + 28.62 ) 132.52
SSblock = − = 2.427
3 15
(v) Calculate the Treatment Sum2 of Squares
40 2 432 49.52 ( y)
= + + − = 9.43
5 5 5 n
(vi) Calculate the Error Sum of Squares
SSerror = SStotal − SSblock − SSregression = 13.69 − 2.427 − 9.43 = 1.833
(vii) Calculate

2.427 9.43
MSblock = = 0.60675 MStreatment = = 4.715
4 2
1.833
MS error = = 0.229
8
4.715
(viii) Calculate F Fblock =
0.60675
= 2.64 Ftreatment = = 20.6
0.229 0.229

SOURCE df SS MS F p
Blocks 4 2.427 0.60675 2.65 not significant
Treatment 2 9.43 4.715 20.6 p<0.001
Error 8 1.833 0.229
Total 14 13.69
Practice questions
1. A Randomised Complete Block Design was used for an experiment on grapes
to test the effect of bird repelling netting on the sugar content of the grapes at
harvest. (The netting has some shading effect, and the beneficial effect of
repelling birds has to be balanced against loss of production.)
There were 8 blocks and 5 netting treatments (one of which is a control: that is
no netting.) Fill out the following part of and an Analysis of Variance Table with
the correct number of degrees of freedom for each of the appropriate sources
of variance. (Note there may be more than two required here.)

Source of variance Degrees of freedom

2. A trial examined the effectiveness of three types of insect traps. The three
traps were set out five separate periods, which can be thought of as “blocks”,
and the average number of insects caught recorded.

The Analysis of Variance table is:

Source of variance Sums of Squares DF Mean Square F ratio P

Trap 427142 2
Period 102715 4 2.64
Error 77924 8 9740.5
Total

What is the Mean Square for the Trap term?

3. In question 2, what is the p-value for the Period F-ratio?

4. In question 2, what is the Standard Error of the Difference (S.E.D.) between

the Trap means?

5. Birth weight of babies born to mothers of various pregnancy weight groups

(taken to be “blocks”) and consumption of cigarettes (none, 1 pack/day, and
more than 1 pack/day) were recorded.

The data and the row and column totals are displayed as follows:
`Rows: weight group Columns: cigarette group
1 2 3 All
1 3.2 2.8 1.7 7.7
2 3.2 2.8 2.5 8.5
3 3.2 3.1 2.5 8.8
4 3.4 3.1 2.6 9.1
5 3.5 3.3 2.8 9.6
6 3.5 3.4 2.9 9.8
All 20.0 18.5 15.0 53.5
(i) Complete the ANOVA table below

Analysis of Variance for birth weight

ANOVA
Source of SS df MS F P-value
Variation
Weight Group 0.983 5 0.1966 5.8 0.01
Cigarette Group (a) 2 (d) (e) 0.0001
Error (b) (c) 0.0339
Total 3.516 17

(ii) Calculate the standard error of the difference between the cigarette
consumption means.

Answers Randomised Complete Block Design

1. Source df working
Block 7 Block df = no. of blocks – 1
=8–1=7
Netting 4 Netting df = no. of netting treatments–1
=5–1=4
Error 28 Error df = Total df – Block df – Netting df
= 39 – 7 – 4 = 28
Total 39 Total df = n – 1
= 40 – 1 = 39

SStrap 427142
2. MStrap = = = 213571
df trap 2

3. P > 0.05.
Method: Look up critical values of F in tables for df1 = 4, and df2 = 8 for
various values of a. F.05,4,8 = 3.84 = 3.838 (depending on your table). Compare
the tables F values with the calculated F (2.64 - from the ANOVA table in the
question). Because Fcalc < F.05, conclude that P > 0.05.

4. 𝑠𝑒𝑑(𝑡𝑟𝑎𝑝𝑠) = √2 × 𝑀𝑆𝐸𝑟𝑟𝑜𝑟 = √2 ×9740 = 62.4195 (No of periods =5)

𝑏 5

(block 2 ) 202 +18.52 +152 53.52

5. (i) (a) SScigarettes = − CF = − = 2.194
k 6 18

(b) SSerror = SStotal − SSweight − SScigarettes = 3.516 − 0.983 − 2.194 = 0.339

2.194
(c) df = 17 – 7 = 10 (d) MScigarette = = 1.097
2
1.097

(e) Fcigarette = = 32.36

0.0339
OR Start with (c) df error = 10 

(b) SS error = 10  MS error = 10  0.0399 = 0.399

 (a) SScig = SStotal − (SSweight + SSerror ) = 3.516 − (0.983 + 0.339) = 2.194

SScig 2.194
 (d) MScig = = = 1.097
df cig 2

 (e) as above.

(ii) 𝑆𝐸𝐷(𝑐𝑖𝑔𝑎𝑟𝑒𝑡𝑡𝑒𝑠) = √2 ×𝑀𝑆𝐸𝑟𝑟𝑜𝑟 = √2 ×0.0339 = 0.1063

𝑏 6
 MODULE 3
LATIN SQUARE DESIGNS (LSD)
A Latin square is a square array of objects (letters A, B, C …) such that each
object appears once and only once in each row and each column.
Example:
A B C D
B C D A
C D A B
D A B C

In LSD, there are three factors to be considered:

1. Rows
2. Columns
3. Treatments (letters A, B, C…)
Latin square design is an experimental design used frequently in agricultural
research. In such a design the treatments are so allocated among the plots that no
treatment occurs more than once in any one row or column.

Preparation of ANOVA table

Source of variation Sum of Degree of Mean sum of squares Variance ratio

squares freedom
𝑆𝑆𝐶 𝑀𝑆𝐶
Between columns SSC K-1 𝑀𝑆𝐶 = 𝐹 =
𝐾−1 𝑀𝑆𝐸
𝑆𝑆𝑅 𝑀𝑆𝑅
Between rows SSR K-1 𝑀𝑆𝑅 = 𝐹 =
𝐾−1 𝑀𝑆𝐸
𝑆𝑆𝐾 𝑀𝑆𝐾
Between treatments SSK K-1 𝑀𝑆𝐾 = 𝐹 =
𝐾−1 𝑀𝑆𝐸
𝑆𝑆𝐸
Residual SSE (K-1) (K-2) 𝑀𝑆𝐸 =
(𝐾 − 1)(𝐾 − 2)
2
Total TSS K -1

Problem :1

The following is a Latin square of a design, when 4 varieties of seeds are being tested.
Set up the analysis of variance table and state your conclusion. You may carry out
suitable change of origin and scale.
1
 A (105) B (95) C (125) D (115)
C (115) D (125) A (105) B (105)
D (115) C (95) B (105) A (115)
B (95) A (135) D (95) C (115)
Solution: Subtract 100 and then divided by 5, we get

A (1) B (-1) C (5) D (3)

C (3) D (5) A (1) B (1)
D (3) C (-1) B (1) A (3)
B (-1) A (7) D (-1) C (3)

𝑋 𝑋 𝑋 𝑋 TOTAL 𝑋 𝑋 𝑋 𝑋
𝑌 1 -1 5 3 8 1 1 25 9
𝑌 3 5 1 1 10 9 25 1 1
𝑌 3 -1 1 3 6 9 1 1 9
𝑌 -1 7 -1 3 8 1 49 1 9
TOTAL 6 10 6 10 T=32 20 76 28 28

𝐻 : There is no significant difference between column means, row means and

treatments.

𝐻 : There is significant difference between column means, row means or

treatments.
STEP 1: Compute the correction factor N
Here, N=16.
STEP 2: Compute the total sum of observations T
T=32
𝑻𝟐
STEP 3: Find
𝑵

= = 64

2
STEP 4: Compute the total sum of squares in columns TSS

TSS= ∑𝑋 +∑𝑋 +∑𝑋 +∑𝑋 -

=20+76+28+28-64 =88
STEP 5: Compute the sum of squares of column SSC
(∑ ) (∑ ) (∑ ) (∑ )
SSC= + + + -

Where , 𝑁 – Is the number of elements in each column.

( ) ( ) ( ) ( )
SSC = + + + – 64

=4
STEP 6: Compute the sum of squares of row SSR
(∑ ) (∑ ) (∑ ) (∑ )
SSR = + + + -

Where , 𝑁 – Is the number of elements in each row.

( ) ( ) ( ) ( )
= + + + – 64

=2

STEP 7: Find sum of squares of treatments SSK

To find SSK, arrange the elements in the order of treatment.

Total
A 1 1 3 7 12
B -1 1 1 -1 0
C 5 3 -1 3 10
D 3 5 3 -1 10

SSK= + + + - 64
=22
STEP 8 : Computing sum of squares of errors SSE
SSE=TSS-SSC-SSR-SSK

3
 =88-4-2-22
=60
STEP 9: ANOVA

Source of Sum of Degree of Mean sum of squares Variance ratio Table F at 5%

variation squares freedom level
𝑆𝑆𝐶
Between SSC=4 K-1=4-1 𝑀𝑆𝐶 =
𝐾−1 MSE>MSC
columns MSE>MSC
= =1.33 𝐹 =
𝑀𝑆𝐸
Fc (6,3)=8.94
𝑀𝑆𝐶

= .
=7.52>1
𝑆𝑆𝑅
Between SSR=2 K-1=4-1 𝑀𝑆𝑅 = MSE>MSR MSE>MSR
𝐾−1 𝑀𝑆𝐸 FR(6,3)=8.94
rows 𝐹 =
= = 0.67 𝑀𝑆𝑅

= .
=14.9>1
𝑆𝑆𝐾
Between SSK=22 K-1=4-1 𝑀𝑆𝐾 = MSE>MSK MSE>MSK
𝐾−1 FK(6,3)=8.94
treatments
= =7.33 𝐹 =
= . =1.36>1

Error SSE=60 (K-1)(K-2)

𝑀𝑆𝐸
𝑆𝑆𝐸
=(4-1)(4-2) =
(𝐾 − 1)(𝐾 − 2)
=6 = = 10

STEP 10: Conclusion

 Calculated FC value < Table FC value, i.e., 7.52 < 8.94.

So, we accept H0 for this case, i.e., there is no significant difference between
columns.

4
  Calculated FR value > Table FR value, i.e., 14.93 > 8.94.

So, we reject Ho for this case, i.e., there are significant differences in rows.

 Calculated FK value < Table FK value, i.e., 1.36<8.94.

So, we accept Ho for this case, i.e., there is no significant difference between
treatments.

5
 FACTORIAL EXPERIMENTS

Two Factorial Experiments

A two-factorial experiment is a study where two factors (variables) are tested simultaneously
to see how they affect an outcome (response). Each factor has two or more levels (e.g.,
high/low or different types). The experiment also examines if the two factors interact,
meaning if their combined effect is different from their individual effects.

Two-factorial experiments study the effects of as on a response variable simultaneously.

These experiments are designed to evaluate the main effects of each factor and their
interaction effects.
Key Terms
1. Factor: An independent variable being tested (e.g., fertilizer type, irrigation level).
2. Level: The different settings or values of a factor (e.g., high/low, organic/chemical).
3. Main Effect: The individual impact of each factor on the response variable.
4. Interaction Effect: How the combination of factors influences the response
differently than their individual effects.

Structure of Two-Factorial Experiments

• Two factors are tested together.
• Each factor has two or more levels.
• All possible combinations of the levels of the two factors are studied.
Example:
Factors:
• Factor A: Fertilizer type (Organic, Chemical).
• Factor B: Irrigation level (High, Low).
Combinations:
• Organic + High
• Organic + Low
• Chemical + High
• Chemical + Low
Steps in a Two-Factorial Experiment
1. Identify Factors and Levels:
o Decide on the two factors and the levels for each.
2. Design the Experiment:
o Combine all factor levels systematically (e.g., full factorial design).
3. Randomize:
o Randomly assign treatments to experimental units to avoid bias.
4. Conduct the Experiment:
o Collect data for all combinations.
5. Analyze Results:
o Calculate the main effects of each factor.
o Determine if an interaction effect exists between the two factors.

Benefits of Two-Factorial Experiments

• Efficiency: Studies two factors simultaneously.
• Interaction Analysis: Identifies if the effect of one factor depends on the level of the
other.
• Comprehensive Insights: Provides information about both individual and combined
effects.

Analysis Using Two-Factorial Experiments

• Use Analysis of Variance (ANOVA):
1. Test the main effects of both factors.
2. Test the interaction effect.
• If the interaction is significant, analyze the combined effects in detail.
Example:
Aircraft primer paints are applied to aluminium surfaces by two methods: dipping and
spraying. The purpose of the primer is to improve paint adhesion, and some parts can be
primed using either application method. The process engineering group responsible for this
operation is interested in learning whether three different primers differ in their adhesion
properties. Examine the influence of each prime type and methods. Is there any interaction
between them?

Methods
Primer type Dipping Spraying
1 4.0,4.5,4.3 5.4,4.9,5.6
2 5.6,4.9,5.4 5.8,6.1,6.3
3 3.8,3.7,4.0 5.5,5.0,5.0

Solution:

Solution:
Define the Hypothesis:
For Primer type:
• H0: μ1=μ2=μ3=0
• H1:at least one μi≠0
For Method type:
• H0: μD=μS=0
• H1:at least one μj≠0
For interaction:
• H0: μ1D1=μ1S1=μ2D1=………=μ3S2=0
• H1:at least one μijk≠0

Primer type Dipping Spraying Total

1 4.0,4.5,4.3 5.4,4.9,5.6 28.7
2 5.6,4.9,5.4 5.8,6.1,6.3 34.1
3 3.8,3.7,4.0 5.5,5.0,5.0 27.0
Total 40.2 49.6 G=89.8
 𝐺2 (89.8)2
Correction Factor, C= =
𝑁 18

SS Total=∑∑∑ x2ijk – C
(89.8)2
= (42)+(4.5)2+……..+(5)2 +(5)2 - 18
=
10.72

𝑇2
SS Primer=∑18𝑖 - C

(28.7)2 (34.1)2 (28.7)2 (89.8)2

= + + −
6 6 6 18

=4.58

𝑇2
SS Method =∑ 𝑛𝑖 - C
𝑗

(40.2)2 (49.6)2 (89.8)2

= + −
9 9 18

= 4.91
𝑇2
SS interaction =∑∑ 𝑛𝑖 - C-SS Primer – SS method
𝑗

(12.8)2 (15.9)2 (11.5)2 (15.9)2 (18.2)2 (15.5)2 (89.8)2

= + + + + + − - 4.58-4.91
3 3 3 3 3 3 18

= 0.24

SS Error= SS Total -SS Interaction

= 0.99

ANOVA Table
Source of Degree of Sum of squares Mean Square F-ratio
variation freedom
Primer 2 4.58 2.29 28.63
Methods 1 4.91 4.91 61.38
Interaction 2 0.24 0.12 1.50
Error 12 0.99 0.08
Total 17 10.72
Critical value F(2,12) (0.05) =3.89
F(1,12)(0.05) =4.75

For Primer types:

Since 28.63 > 3.89; thus H0 is rejected
For methods:
Since 61.38 >4.75 ; thus H0 is rejected
For interaction:
Since 1.50 < 3.89 ; thus H0 is accepted
Thus there is no interaction between primer types and methods.