CHAPTER 5

FACTORIAL DESIGNS
(CONT.)
Example: (5-1)
An experiment is conducted to determine
the effects of the plate material and
temperature on the life of a battery. Three
levels of each factor selected, and a factorial
experiment with four replicates is
performed. The life data (in hours) follow:
Material Temperature oF (B)
Type (A) 15 70 125
1 130 155 34 40 20 70
74 180 80 75 82 58

2 150 188 136 122 25 70

159 126 106 115 58 45

3 138 110 174 120 96 104

168 160 150 139 82 60
The statistical Model:
Observed values of the response yijk can
be represented by the linear model:-
yijk     i   j  ( ) ij   ijk
Where,
(ij is a parameter representing the
average effect of the interaction between
the ith row and the jth column.
The hypotheses of interest:

1. Equality of row effects:

Ho: i = 0 Vs. H1: i  0
2. Equality of column effects:
Ho: j = 0 Vs. H1: j  0
3. No interaction effects:
Ho: (ij = 0 Vs. H1:
(ij  0
The ANOVA Proposition
The total variability in observed responses
can be decomposed into four components:
• Variability due to row effects.
• Variability due to column effects,
• Variability due to row-column
interactions, and
• Variability due to experimental Error.
The ANOVA Proposition
In terms of the Sum of Squares:
SSTotal = SSA+ SSB+ SSAB+SSError

The corresponding decomposition of the

number of degrees of freedom is:
(N-1) = (a-1) + (b-1)+ (a-1)(b-1)+ab(n-1)
Where,
a b n 2
y
SSTotal   yijk 2
 ...
i 1 j 1 k 1 abn
Effects,
1 a 2 y...2
SS A  [  yi.. ] 
bn 1 abn
1 b 2 y...2
SS B  [  y. j . ] 
an 1 abn

1 a b 2 y...2
SS AB  [  yij . ]   SS A  SS B
n i 1 j 1 abn
ANOVA Table:

Source SS Df MS F0 P-value

A 10683.72 2 5341.86 7.91 0.0020

B 39118.72 2 19559.36 28.97 0.0001
AB 9613.78 4 2403.44 3.56 0.0186
Error 18230.75 27 675.21

Total 77646.97 35

There is a significant interaction between the

type of material (A) and temperature level
(B), at  > 0.0186.
Model Adequacy Checking
To detect violations of the basic assumptions
of the ANOVA procedure, we need to
examine the residuals given by:

eijk  yijk  yij .

Where,yij. is the cell average
The battery life Example:
1
y ..1  [130  155  74  180]  134.75
4
e111  130  134.75  4.75
e112  155  134.75  20.25
e113  74  134.75  60.75
e114  180  134.75  45.25

See complete Table in the book

NOPP of residuals:

99
Normal % Probability
95
90
80
70
50
30
20
10
5
1

-60.75 -34.25 -7.75 18.75 45.25

Residual
Residuals vs. Fitted values:

45.25

18.75
Residuals

-7.75

-34.25

-60.75

49.50 76.06 102.62 129.19 155.75

Predicted (Cell Average)

Residuals vs. Material (A):

45.25

18.75
Residuals

-7.75

-34.25

-60.75

1 2 3

Material
Residuals v. Temp. (B):

45.25

18.75
Residuals

-7.75

-34.25

-60.75
1 2 3
Temp
Residuals v. Run Order:

Run order not given!

>>No severe violations are found<<

Interaction Plot

160
Material (A)
140 1
2
Average Life

120
3
100
80
60
40
15 70 125

Temperature level (B)

Interpretation:
• At the low temperature level, the average
battery life is high regardless of the type of
material used.
• At the intermediate temperature level,
material type 3 shows the highest average
life.
• At the high temperature level, all averages
decrease with material type 3 having the
highest average life.
Conclusion:
The use of material type 3 is recommended
since it results in a minimum change of the
average life over all temperature levels.

Note:
Interpretations are given in terms of the
main effects when the interaction effect is
not significant.
Estimation of Parameters:
ˆ  y...
ˆi  y i..  y...
ˆ j  y. j .  y...
( ) ij  y ij .  y i..  y. j .  y...
The General Factorial design:
An extension of the two factor factorial
with a levels of A, b levels of B, c levels of
C, and so on.

Assumptions:
• Independent factors,
• Two or more replicates n > 2
• Random order
The statistical model:
For a three-factor factorial:
y ijk     i   j   k
 ( ) ij  ( ) ik  (  ) jk
 ( ) ijk   ijk
Hypotheses of interest:
H o :  i  0 vs. H 1 : at least one i  0
H o :  j  0 vs. H 1 : at least one  j  0
H o :  k  0 vs. H 1 : at least one  k  0
H o : ( ) ij  0 vs. H 1 : at least one ( ) ij  0
H o : ( ) ik  0 vs. H 1 : at least one ( ) ik  0
H o : (  ) jk  0 vs. H 1 : at least one (  ) jk  0
H o : ( ) ijk  0 vs. H 1 : at least one ( ) ijk  0
The Sum of Squares
SS Total  SS A  SS B  SS C
 SS AB  SS AC  SS BC
 SS ABC  SS Error
Degrees of Freedom
abcn  1  (a  1)  (b  1)  (c  1)
 (a  1)(b  1)  (a  1)(c  1)  ((b  1)(c  1)
 (a  1)(b  1)(c  1)  abc(n  1)
Where,
2
a b c n y
SS Total      y 2
ijkl  ....
1 1 1 1 abcn
Main Effects,
2
1 a y
SS A   y i... 
2 ....

bcn 1 abcn
2
1 b y
SS B   y. j .. 
2 ....

acn 1 abcn

1 c 2 y....2
SS C   y..k . 
abn 1 abcn
Two-Factor Interactions,
2
1 a b y....
SS AB  [   y ij .. 
2
]  SS A  SS B
cn 1 1 abcn
1 a c 2 y....2
SS AC  [   y i.k .  ]  SS A  SS C
bn 1 1 abcn
2
1 b c y....
SS BC  [   y. jk . 
2
]  SS B  SS C
an 1 1 abcn

With [.] represents SSsubtotals( . )

Three-Factor Interactions,
2
1 a b c y
SS ABC  [    y ijk . 
2 ....
]  SS A  SS B  SS C
n 1 1 1 abcn
 SS AB  SS AC  SS BC

The Error Term,

SS Error  SS Total  SS subtotals ( ABC )

Example: (5-3) P.184
A study is conducted to determine the effect
of three variables on a filling process. The
percent carbonation (A) is tested at three
levels, operating pressure (B), and line
speed (C) both at two levels. The
following data represent the results in terms
of the average deviation from the target fill
height (response):
 Pressure (B)
25 psi 30 psi
Carbonation Speed (C) Speed (C)
(A)% 200 250 200 250
10 -3 -1 -1 1
-1 0 0 1
12 0 2 2 6
1 1 3 5
14 5 7 7 10
4 6 9 11
The ANOVA Table:

Source SS Df MS F0 P-value

A 252.75 2 126.375 178.412 0.0000

B 45.375 1 45.375 64.059 0.0000
C 22.042 1 22.042 31.118 0.0001
AB 5.250 2 2.625 3.706 0.0558
AC 0.583 2 0.292 0.412 0.6713
BC 1.042 1 1.042 1.471 0.2485
ABC 1.083 2 0.542 0.765 0.4867
Error 8.500 12 0.708

Total 336.625 23
Model Adequacy
Residuals are given by:
eijkl  y ijkl  y ijk .
Examine:
• NOPP of residuals
• Residuals vs. fitted values
• Residuals vs. factor levels
• Residuals vs. run order (if given)
No severe violations are found, and we
conclude that there is a significant interaction
between carbonation (A) and pressure (B)
At >0.0558.
 AC Interaction Plot
8.7 Factor_C
Average Deviation

6.7 1
4.7 2
2.7
0.7
-1.3
10 12 14

Carbonation

No significant interaction.
 BC Interaction Plot
Factor_C
8.7
Average Deviation

6.7 1
4.7 2
2.7
0.7
-1.3
25 30

Pressure

No significant interaction.
 AB Interaction Plot
Average Deviation Factor_A
10
8 1
6 2
4 3
2
0
-2
1 2
Pressure
At low pressure, the Carbonation percentage
has lower effect on the average deviation.
 Main Effect Plot
Average Deviation 4.6
4.1 4.083
3.6
3.1 3.125
2.6
2.1 2.167
1.6
1 2
Line Speed

Operating at a line speed of 200 bpm

reduces the average deviation by 0.958 units.
Response Curves & Surfaces:
The Regression model of the response (y)
and the levels of experimental factors used
to help interpret, and implement results.

Conditions:
• Two or more quantitative factors,
• Interpolation purposes only.
 Estimated Response Surface
Fill Deviation Speed=1.0

11
9
7
5
3
1 1
-1 0
-1 0 -1
1
Pressure
Carbonation
 Contours of Estimated Response Surface
Factor_C=1.0
1
Pressure

-1
-1 0 1

Carbonation