Chapter 5 Design of Experiment

Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 5
Chapter - Five
Factorial Experiment
Contents
5. Factorial Designs
5.1 Basic definitions and principles
5.2 The advantage of factorial designs
5.3 The two factor factorial design
5.3.1 Two factor design without interaction
5.3.2 Two factor design with interaction
5.4 The three factor factorial design
5.5 Blocking in a factorial design
5.6 Unbalanced data in a factorial design
5.7 The 2k factorial design
5.7.1 The 22 design
5.7.2 The 23 design
5.8 Blocking and confounding in 22 and 23 designs
Factorial Designs
“Do not put your faith in what statistics say until you have carefully
considered what they do not say”. W. WATT
A factorial design is the one in which two or more factors are investigated simultaneously in
the same study. In other words, factorial design involves testing two or more different factors on
the same group at the same time (in one period).
natnaelmoges19@gmail.com
108
Although single - factor designs are useful for illustrating the basic mathematics of analysis of
variance; they are probably from a minority of actual research in every scientific discipline. The
majority of ANOVA analysis reported in journals have at least two and three factors. This is
because, these days, single-factor designs are often inadequate for shedding light upon the
complicated research questions that every professional are trying to answer. Therefore, this
chapter is designed to introduce the basic types of factorial designs with the corresponding
statistical analysis.
5.1 Basic definitions and principles
In any types of factorial design, we have two important terms that are more frequently
mentioned.
Main effect is the effect of independent variable (factor) on a dependent variable (response
variable) averaging across the levels of any other independent variables.
Interaction may arise when considering the relationship among three (one for response
variable and the other two for factors) or more variables, and describes a situation in which the
simultaneous influence of two variables on a third is not additive1.
The presence of interactions can have important implications for the interpretation of statistical
models. If two variable of interest interact, the relationship between each of the interacting
variables and a third, response variable, depends on the value of the other interacting variable.
In practice, this makes it difficult to predict the consequence of changing the value of a variable,
particularly if the variables it interacts with are hard to measure or difficult to control.
Every factorial design models may not contain interaction term. However, the presence or the
absence of the interaction term in the model is determined by some graphical techniques such as
response surface and contour plot. Response surface becomes curvature in the existence of
interaction whereas there is no interaction. In case of contour plot, the lines become parallel if
there is no interaction.
109
1
In number theory, an additive function is an arithmetic function of positive integer such that
whenever and are cop rimes, the function of the product is the sum of the functions:
5.2 The advantage of factorial designs
As we know in chapter one, factorial experiment is one of the strategies of experimentation. A

design which uses factorial experiment strategy is called factorial design. Prominent text books
and academic papers currently favor factorial experimental designs, a method pioneered by Sir
Ronald A. Fisher, where multiple factors are changed at once.
The Advantage of Factorial designs
Factorial designs have several advantages some of them are:

 They are more efficient than one – factor – at – a - time experiments.
 Factorial design is necessary when interactions may be present to avoid misleading
conclusions.
 Factorial designs allow the effects of a factor to be estimated at several levels of the other
factors, yielding conclusions that are valid over a range of experimental conditions.
 Therefore, the reasons stated for favoring the use of factorial design over one-factor-
at - a time (OFAT) are:
OFAT requires more runs for the same precision in effect estimation.
OFAT cannot estimate interactions.
OFAT can miss optimal settings of factors.
110
5.3 The two factor factorial design
The simplest factorial design which serves as a basis for the establishment of the other types of
advanced factorial designs (i.e., higher order factorial design such as three factor, four factor
factorial designs etc.) is two-factor factorial design. This design only considers two-factors and
according to the existence of interaction effect it may consider the interaction of two-factors. As
a result, we will have two versions of two-factor factorial design models: with interaction and
without interaction. Assume the number of observations in each cell are equal (i.e., balanced
data).
Let be the observed response when factor A is at the ith level (i = 1, 2, . . . , a) and factor B is
at the jth level ( j =1, 2, . . . , b) for the kth replicate (k =1, 2, . . . , n). In general, a two-factor
factorial design has abn observations which are taken randomly so that it can also be called as
Completely Randomized two-factor factorial design.
Then the layout of data looks like the following:
Factor B
1 2 … Total
1
… = … = … … =
2
Factor A
… = … = … … =
…
… = … = … … =
Total …
111
In the two-factor factorial, both row and column factors (and treatments), Say A and B, are of
equal interest.
Specifically, we are interested in testing hypothesis about the equality of row treatment effects,
say
and the equality of column treatment effects, say
We are also interested in determining whether row and column treatments interact. Thus, we also
wish to test
5.3.1 Two factor design without interaction
When someone thinks that there is no interaction effect, the two - factor factorial design model
will be
Where is the overall mean effect, is the effect of the level of the row factor A, is the
effect of the level of the column factor B and is a random error component. Both factors
are assumed to be fixed and the treatment effects are defined as deviations from the overall mean,
so Σ = 0 and Σ = 0.
112
Let denotes the total of all observations under the level of factor A, denotes the total
of all observations under level of factor B, denotes the total of all observations in the
cell and denotes the grand total of all observations.
Let ̅ ̅ ̅ ̅ be defined as the corresponding row, column, cell and grand averages,
respectively. We express mathematically.
∑∑ ̅ ∑ ̅
∑∑ ̅ ∑∑∑ ̅
The total corrected sum of squares can be rewritten as
∑ ∑ ∑( ̅ ) ∑ ∑ ∑{ ̅ ̅ (̅ ̅ ) ( ̅ ̅ ̅ )}
∑ ̅ ̅ ∑ ̅ ̅ ∑ ∑ ∑( ̅ ̅ ̅)
That is, .
 For a sake of simplifying computation of sum of squares
∑∑∑ ∑
113
Then the ANOVA table for two - factor factorial design without interaction is
Source of variation Sum of squares DF Mean square
Row(A) a-1
Column(B) b-1
Error ab(n-1)
Total
Occasionally, one encounters a two - factor experiment with only a single replicate, that is, only
one observation per cell. Then the sum squares must be arranged in the following way:
∑∑ ∑
The corresponding degrees of freedom are
o The appropriate mode is then
Note that: If the two-factor factorial design has a single observation per a cell, then there is no
interaction term in the model.
114
5.3.2 Two factor design with interaction
Now, let us assume that there is interaction effect and the appropriate mode will be
Where is the overall mean effect, is the effect of the ith level of the row factor A, is the
effect of the jth level of column factor B, is the effect of the interaction between and ,
and is a random error component. Both factors are assumed to be fixed, and the treatment
effects are defined as deviations from the overall mean, so ∑ and ∑ .
Similarly, the interaction effects are fixed and are defined such that ∑ ∑ .
Because there are n replicates of the experiment, there are abn total observations.
∑ ∑ ∑{ ̅ ̅ (̅ ̅ ) (̅ ̅ ̅ ̅) ( ̅ )}
∑ ̅ ̅ ∑(̅ ̅ ) ∑ ∑( ̅ ̅ ̅ ̅ ) ∑ ∑ ∑( ̅ )
For simplicity of computation,
∑∑∑ ∑
115
It is convenient to obtain the in two stages. First, we compute the sum of squares between
the totals, which is called the sum of squares due to subtotals:
∑∑
This sum of squares also contains and . Therefore, the second step is to compute
as:
[ ∑∑ ]
We may also compute by subtraction as follows:
The ANOVA table looks like;
Row(A) a-1
Column(B) b-1
Interaction(AB) (a-1)(b-1)
Error ab(n-1)
Total
NB: DF: Degree of Freedom
116
Example. The effective life (in hours) observed in the battery design is studied. The two most
important variables that are thought to effect life of battery design are the temperature and
material type. Three levels of each factor is selected with four replication were performed. The
data is as follows.
Then based on the above information, Test the significance of temperature, material type and
their interaction at 5% level of significance and perform anova table.
Solution : a = 3 , b = 3, n = 4
117
= = = 5341.86 = = 7.91
= = = 19559.36 = = 28.97
= = = 2403.44 = = 3.56
= = = = 675.21
Step 1: State the null hypothesis and alternative hypothesis.
i, for factor A (for Material type)
Vs
ii, for factor B ( for Temperature)
Vs
iii, for interaction ( AB (Material*Temperature) )
Vs
Step 2: Identify the level of significance, .
Step 3: Identify the test statistic, F – test.
118
Step 4: Determine test statistic.
= = 7.91
= = 28.97
= = 3.56
Step 5: Identify the critical value
= = = 3.35,for Material type

= = = 3.35, for Temperature
= = = 2.73,
for Interaction (Material*Temperature)
Step 6: Make decision:
o Reject if , , for Material type.
o Reject if , ) for Temperature.
o Reject if , for interaction.
 7.91 & 28.97 > 3.35, so reject for Material type and Temperature.
 3.56 > 2.73 so reject for interaction (Material*Temperature).
Step 7: Conclusion:
At 5% level of significance we have enough evidence to conclude that
there is significance difference among Material type.
there is significance difference among Temperature.
there is significance difference among interaction (At 5% level of
significance we have enough evidence to conclude that there is a
significant interaction between material type and temperature).
119
The ANOVA table becomes:
Material Type a-1
Temperature b-1
Interaction (a-1)(b-1)
Error ab(n-1)
Total
Material Type 10683.72 2 5341.86 7.91

Temperature 39118.72 2 19559.36 28.97
Interaction 9613.78 4 2403.44 3.56
Error 18230.75 27 675.21
Total 77646.97
5.4 The three factor factorial design
So far, we have seen the ANOVA for two-factor factorial design. We can easily extend the
principle of two-factor factorial design for three factors. Let, A, B and C are the three factors
with the corresponding levels , and , respectively.
A three-factor analysis of variance consists of seven significance tests: a test for each of the three
main effects, a test for each of the three two-way interactions, and a test of the three-way
interaction.
The model for three factor factorial design is
120
Where is the observation of the level of factor and level of factor in the
level of factor , is the overall mean, is the effect of the ith level of the row factor A, is
the effect of the jth level of column factor B, , is the two-way interaction effect of level
of factor and the level of factor , is the two-way interaction effect of level of
factor and the level of factor , is the two-way interaction effect of level of
factor and the level of factor , is the three-way interaction of the level of
factor , level of factor and level of factor and is the usual random disturbance
term.
In this analysis, the total variation can be partitioned (decomposed) into eight sums of squares.
That is,
∑∑∑∑ ∑
∑ ∑
[ ∑∑ ]
[ ∑∑ ]
121
[ ∑∑ ]
[ ∑∑∑ ]
122
Then the ANOVA table will look like the following:
Source of Variation Sum SQUARE DF Mean Square

A (a-1) /
B (b-1) /
C (c-1) /
AB (a-1)(b-1) /
AC (a-1)(c-1) /
BC (b-1)(c-1) /
ABC (a-1)(b-1)(c-1) /
Error abc(n-1)
Total abcn-1
Using this ANOVA, we can test the following seven hypothesis. These are:
 Reject the null hypothesis if / > { }
123
Example:-
A soft drink bottler is interested in obtaining more uniform fill heights in the bottles produced by his
manufacturing process. The filling machine theoretically fills each bottle to the correct target height, but
in practice, there is variation around this target, and the bottler would like to understand better the sources
of this variability and eventually reduce it.
The process engineer can control three variables during the filling process: the percent
carbonation (A), the operating pressure in the filler (B), and the bottles produced per minute or
the line speed (C). The pressure and speed are easy to control, but the percent carbonation is
more difficult to control during actual manufacturing because it varies with product temperature.
However, for purposes of an experiment, the engineer can control carbonation at three levels: 10,
12, and 14 pecent.She chooses two levels for pressure (25 and 30 psi) and two levels for line
speed (200 and 250 bpm). She decides to run two replicates of a factorial design in these three
factors, with fall 24 runs taken in random order. The response variable observed is the average
deviation from the target fill height observed in a production run of bottles at each set of
conditions. The data that resulted from this experiment is shown below.
124
Based on the above information: Test the significance of all factors, and their
interactions at 5% level of significance and perform anova table.
Solution: a = 3, b = 2, c = 2, n = 2
The total corrected sum square is found as:
125
The sums of squares for the main effects are calculated as:
The sum of squares for the two – factor interactions are calculated as:
o To find the carbonation- pressure or AB interaction, we need the totals

for the A { }.
126
o To find the carbonation- speed or AC interaction, we need the totals for

the A { }.
o To find the pressure - speed or BC interaction, we need the totals for the
B { }.
o To find the carbonation, pressure and speed or ABC interaction, we need

the totals for the A B { }.
127
= = = 126.375 = = 178.412
= = = = = 64.059
= = = = = 31.118
= = = 2.625 = = 3.706
= = = 0.292 = = 0.412
= = = 1.042 = = 1.471
= = = 0.542 = = 0.765
= = = = 0.708
i, for factor A (for carbonation)
VS
ii, for factor B (for pressure)
VS
iii, for factor C (for speed)
VS
iv, for factor AB (for carbonation and pressure)
VS
v, for factor AC (for carbonation and speed)
VS
vi, for factor BC (for pressure and speed)
VS
vii, for factor ABC (for carbonation, pressure and speed)
VS
128
= = 178.412 = = 64.059
= = 31.118 = = 3.706
= = 0.412 = = 1.471 = = 0.765
 { }= = 3.89, for factor A.

 { }= = 4.75, for factor B.
 { }= = 4.75, for factor C.
 { }= = 3.89, for factor AB.
 { }= = 3.89, for factor AC.
 { }= = 4.75, for factor BC.
 { }= = 3.89, for factor ABC.
o Reject if , critical value for all.
 178.412 > 3.89, 64.059 > 4.75, 31.118 > 4.75, so reject for factor: A, B & C
respectively.
 3.706 < 3.89, 0.412 < 3.89, 1.471 < 4.75 & 0.765 < 3.89, so don’t reject for
factor interaction: AB, AC, BC & ABC respectively.
Step 7: Conclusion:
At 5% level of significance we have enough evidence to conclude that the
percentage of carbonation; operating pressure and line speed
significantly affect the fill volume (At 5% level of significance we have
sufficient evidence to conclude that there is significant difference among
the percentage of carbonation; operating pressure and line speed).
129
At 5% level of significance we have no enough evidence to conclude that
there is a significant interaction between percentages of carbonation
with operating pressure, percentage of carbonation with line speed,
operating pressure with line speed, percentage of carbonation with
operating pressure with line speed.
The ANOVA table looks like the following:
Source of Variation Sum DF Mean

Square Square
Percentage of carbonation(A) (a-1) /
Operating pressure(B) (b-1) /
Line speed(C) (c-1) /
AB (a-1)(b-1) /
AC (a-1)(c-1) /
BC (b-1)(c-1) /
ABC (a-1)(b-1)(c-1) /
Error abc(n-1)
Total abcn-1

Square Square
Percentage of carbonation(A) 252.750 2 126.375
Operating pressure(B) 45.375 1 45.375
Line speed(C) 22.042 1 22.042
AB 2.250 2 2.625 3.706
AC 0.583 2 0.292
BC 1.042 1 1.042 1.471
ABC 1.083 2 0.542 0.765
Error 8.500 12 0.708
Total 336.625 23
NB: DF: degree of freedom.
130
5.5 Blocking in a factorial design
Up to now, we have discussed factorial designs in the context of a completely randomized

experiment. Sometimes, it is not feasible or practical to completely randomize all of the runs in a
factorial. The presence of nuisance factor, for instance, may require the experiment be run in
blocks. We know that a factorial experiment with two - factors (A and B) and n replicates has a
linear statistical model.
The above model is functional as long as the experiment doesn’t need blocking. If the
experiment needs blocking, then the appropriate statistical model will be:
Where is the effect of the block.

For two-factor factorial in randomized complete block, the total sum of squares (total
Variability) can be decomposed into block, A, B, AB and error sum of squares. That is,
∑∑∑ ∑
∑ ∑
∑∑
131
[ ∑∑ ]
Blocks n-1
A a-1
B b-1
Error (ab-1)(n-1)
Total
Note that: The tests are the same as two-factor factorial design with interaction. However, the
degree of freedom for error is different. As a result, it needs due attention while you find
tabulated values.
132
Example:
An engineer is studying methods for improving the ability to detect targets on a radar scope, two
factors she considers to be important are the amount of background noise, or “ground clutter,”
on the scope and the type of filter placed over the screen. An experiment is designed using three
levels of ground clutter and two filter types. We will consider these as fixed type factors. The
experiment is performed by randomly selecting a treatment combination (ground clutter level
and filter type) and then introducing a signal representing the target in to the scope. The intensity
of this target is increased until the operator observes it.
The intensity level at detection is then measured as the response variable. Because of operator
availability, it is convenient to select an operator and keep him or her at the scope until all the
necessary runs have been made. Furthermore operators differ in their skill and ability to use the
scope. Consequently, it seems logical to use the operators as blocks. Four operators are randomly
selected .Once an operator is, chosen the order in which the six treatment combinations are run is
randomly determined.
The Data are shown below
Test the significance of the filter type, Ground clutter , their interaction at 5%
level of significance and perform anova table.
Solution: , abn = 3*2*4 = 24, = 2278(the sum of all observations)
133
∑∑∑
[ ]
∑
[ ]
∑
( )
∑
∑∑
i, for factor A(Ground clutter)
VS
ii, for factor B (Filter type)
VS
iii, for factor AB
VS
134
 { }= = 3.68, for factor A.

 { }= = 4.54, for factor B.
 { }= = 3.68, for factor AB.
 15.73 > 3.68, 96.19 > 4.54, so reject for factor A, factor B respectively.
 3.48 < 68, so don’t reject for factor AB.
Step 7: Conclusion:
ground clutter level is significant (ground clutter level were used affect
the operator’s ability to detect the target).
type of filter is significant (type of scope filter were used affect the
operator’s ability to detect the target).
there is a significant interaction between ground clutter and filter type.
135
Blocks n-1
A a-1
B b-1
Error (ab-1)(n-1)
Total
Blocks 402.17 3 134.06

A(ground clutter) 335.58 2 167.79 15.13
B(filter type) 1066.67 1 1066.67 96.19
Interaction(AB) 77.08 2 38.54 3.48
Error 166.33 15 11.09
Total 2047.83 23
136
5.6 Unbalanced data in a factorial design
So far we have computed ANOVA when each cell contains the same number of responses, say n.
In this section, we will try to make a bit adjustment on the sum of squares.
Let us begin with two-factor factorial design with interaction.
The model will be
cell.
∑∑∑ ∑
∑ ∑ ∑ ∑
∑ ∑
∑ ∑ ∑ ∑
Row(A) a-1
Column(B) b-1
Error Subtraction
Total
∑∑
137
For three factor factorial design, the model will be
cell in the
∑∑∑ ∑ ∑
∑ ∑ ∑ ∑ ∑ ∑
∑ ∑
∑ ∑ ∑ ∑ ∑ ∑
[ ∑∑ ]
∑ ∑ ∑
[ ∑∑ ]
∑ ∑ ∑
[ ∑∑ ]
∑ ∑ ∑
138
[∑ ∑ ∑ ]
∑ ∑ ∑
The degrees of freedom for the main effects and interaction effect are the same with the balanced
dataset. But the degree of freedom for is is(∑ ∑ ∑ ) and you can obtain the
error degrees of freedom using the usual technique.
139
5.7 The 2k factorial design
More practically, factorial designs are widely used. This is due to having interest of considering
several factors and, moreover, they enable the researcher to see the joint effect of the factors on
the response variable. In the previous topics of this chapter, we have considered the two or three
factor factorial design with different levels. Since there are special types of factorial design, we
need to give serious attention or consideration.
The most important of these types is that of factors, each at only two levels. These levels may
be quantitative or qualitative. A complete replicate of such a design requires 2 × 2 × ×2=
observations and is called a factorial design. The base is the number of levels and the power is
for the number of factors. As a result, the next two subtopics deal with extremely important types
of factorial design by keeping in mind that factors are fixed, designs are completely
randomized and the usually normality assumptions are satisfied.
 The factorial design: refers to a special case of the general factorial design in
which there are k factors, each at two levels.
 The two levels are usually called low and high (could be either quantitative or
qualitative).
 Assumptions: (1) the factor is fixed, (2) the design is completely randomized and (3) the
usual normality assumptions are satisfied
The factorial design is particularly useful in the early stages of experimental work, when
there are likely to be many factors to be investigated. It provides the smallest number of runs
with which k factors can be studied in a complete factorial design.
140
5.7.1 The 22 design

The first design in the s is one with only two factors, say A and B, each run at two
levels. This design is called a factorial design The levels of the factors may be arbitrary
called “low” and “high”.
Let low be denoted by − and high denoted by + . From our previous discussion of two - factor
factorial design, the layout of the data will look like.
Factor B
Low High Total
Low
. .
. .
. .
Factor A
High
. .
. .
. .
Total
So as to make it easily accessible (suitable) for the currently undergoing discussion, let us change
the above table into the following format
Factors Treatment
A B Combination Replication Total
A low, B low , ,…,
A high, B low , ,…,
A low, B high , ,…,
, ,…,
A high, B high
141
Note that:
Conventionally, we denote the effect of a factor by a capital letter. Thus, A refers to the
effect of factor A, B refers to the effect of factor B and AB refers to the AB interaction.
The four treatment combinations in the design are also represented by lowercase letters.
Thus, represents the treatment combination of at the high level and at low level,
represents at low level and at high level, represents both factors at high level. By
convention, (1) is used to denote both factors at the low level.
In a two - level factorial design, we may define the average effect of a factor as the
change in response produced by a change in the level of that factor averaged over the
level of the other factor.
Now the effect of at the low level of is and the effect of at the high level
of is . Averaging these two quantities yields the main effect of A.
{ }
( ) { }
{ }
We define the interaction effect AB as the average difference between the effect of A at the high
level of B and the effect of A at the low level of B. Thus,
{ }
142
The sign and the magnitude of the factors determine which variables are likely to be important.
For instance, if the effect of is positive, then this suggests that increasing from low level to
high level will increase the value of the response variable. If the effect of is negative, then this
suggests that increasing from low level to high level will decrease the value of the response
variable.
 Just to find the sum of squares for , and , let us consider the following contrasts.
We usually call these the total effect of , and , respectively. Moreover, these contrasts are
orthogonal (refer to Chapter 3). Therefore,
The total sum of squares is found in the usual way, that is,
∑∑∑
In general,
, is usually computed by subtraction as
143
It is often convenient to write down the treatment combinations in the order (1), , , . This is
referred to as standard order. Using this standard order, we can see the coefficients’ of contrast
used in the estimation of the effect.
Effects (1)
Or equivalently it can be expressed in the following form

Treatment Factorial effect
Combination I A B AB
(1)
A 1
B 1
AB 1
Error 4(n 1)
Total
Example:
Consider an investigation into the effect of the concentration of the rectant and the amount of the
catalyst on the conversion (yield) in a chemical process. Let the reactant concentration be factor
A , and let the two levels of interest be 15 and 25 percent.The catalyst is factor B , with the high
level denoting the use of 2 pounds of the catalyst and the low level denoting the use of only 1
pound. The experiment is replicated three times, and the data are as follows :
144
Test the significance of the main effects(reactant conccentration and catalyst)
and their interaction at 5% level of significance and perform anova table.
Solution:
80, , ,n=3,
∑∑∑
= 9398.00
145
= = = 208.33 = = 53.15
= = = 75.00 = = 19.13
= = = 8.33 = = 2.13
= = = = 3.92
i, for factor A ( for reactant concentration )
Vs
ii, for factor B ( for catalyst)
Vs
iii, for interaction ( AB (reactant concentration * catalyst)) )
Vs
= = 53.15
= = 19.13
= = 2.13
= = = 5.32 , for factor, reactant
concentration(A) , catalyst (B) & for interaction(AB).
146
o Reject if , for factor A , B & AB.
 53.15 & 19.13 > 5.32, so reject for reactant concentration, for
catalyst.
 2.13 < 5.32 so don’t reject for interaction (A*B).
Step 7: Conclusion:
main effects (reactant concentration & catalyst) are statistically
significant.
there is statistically significant difference among the interaction(There
is no interaction between these factors at 5% level of significance).
The ANOVA table becomes:
A 1
B 1
AB 1
Error 4(n 1)
Total
A 208.33 1 208.33 53.15

B 75.00 1 75.00 19.13
AB 8.33 1 8.33 2.13
Error 31.34 8 3.92
Total 323.00
147
5.7.2 The 23 design

Now suppose we have three factors, say A, B and C, each at two levels, are of our interest. The
design is called a factorial design and the eight treatment combinations of the standard order
will be presented below:
Treatment Factorial effect

Combination I A B AB C AC BC ABC
(1)
Then the average effects can be computed as follows:
148
With the same technique as of factorial design contrasts,
and the same technique holds fine for the rest effects. The sum square for factor is then
In general, for factorial designs with n replicates, the sum of squares for any effect is
149

Square Square
A 1 /
B 1 /
C 1 /
AB 1 /
AC 1 /
BC 1 /
ABC 1 /
Error 8(n-1)
Total 8n-1
Example: A study on the effect of percentage of carbonation, operating
pressure, and line speed on the fill height of a carbonated beverage. Suppose
that only two levels of carbonation are used so that the experiment is a
Factorial design with two replicates. The data, deviations from the target fill
height, are shown below.
Test the significance of the main effects (percentage of carbonation, operating
pressure, and line speed) and their interaction at 5% level of significance and
perform anova table.
150
Solution:
151
The sum of squares can be calculated as:
 Generally sum of squares can be calculated as follows:
152
In general,
, is usually computed by subtraction.
= = = 36.00 = = 57.60
= = = = = 32.40
= = = = = 19.60
= = = 2.25 = = 3.60
= = = 0.25 = = 0.40
= = = 1.00 = = 1.60
153
= = = 1.00 = = 1.60
= = = = 0.625
i, for factor A ( for percentage of carbonation )
VS
ii, for factor B (for operating pressure)
VS
iii, for factor C (for line speed)
VS
iv, for factor AB (for percentage of carbonation and pressure)
AB is not significantVS
v, for factor AC (for percentage of carbonation and line speed)
AC is not significant VS
vi, for factor BC (for operating pressure and line speed)
BC is not significant VS
vii, for factor ABC (for percentage of carbonation, operating pressure and line
speed)
ABC is not significant VS
154
= = 57.60 = = 32.40
= = 19.60 = = 3.60
= = 0.40 = = 1.60 = = 1.60
 { }= = 5.32, for factor A.

 { }= = 5.32, for factor B.
 { }= = 5.32, for factor C.
 { }= = 5.32, for factor AB.
 { }= = 5.32, for factor AC.
 { }= = 5.32, for factor BC.
 { }= = 5.32, for factor ABC.
 57.60 > 5.32, 19.60 > 5.32, 32.40 > 5.32, so reject for factor: A, B & C
respectively.
 3.60 < 5.32, 0.40 < 5.32, 1.60 < 5.32 & 1.60 < 5.32, so don’t reject for factor
interaction: AB, AC, BC & ABC respectively.
Step 7: Conclusion:
percentage of carbonation; operating pressure and line speed ARE
significant (At 5% level of significance we have sufficient evidence to
conclude that there is significant difference among the percentage of
carbonation; operating pressure and line speed).
155
there is a significant interaction between percentage of carbonation with
operating pressure, percentage of carbonation with line speed , operating
pressure with line speed , percentage of carbonation & operating pressure
& line speed.

Square Square
Percentage of carbonation(A) 1 /
Operating pressure(B) 1 /
Line speed(C) 1 /
AB 1 /
AC 1 /
BC 1 /
ABC 1 /
Error 8(n-1)
Total 8n-1
NB: n = 2

Square Square
Percentage of carbonation(A) 36.00 1 36.00 57.60
Operating pressure(B) 20.25 1 20.25 32.40
Line speed(C) 12.25 1 12.25 19.60
AB 2.25 1 2.25 3.60
AC 0.25 1 0.25 0.40
BC 1.00 1 1.00 1.60
ABC 1.00 1 1.00 1.60
Error 5.00 8 0.625
Total 78.00 15
NB: DF: degree of freedom.
156
5.8 Blocking and confounding in 22 and 23 designs
In replicated designs where we have n replications per cell and perform a completely randomized
design we randomly assign all times n experimental units to the 2k treatment combinations.
Alternatively, when we have n replicates we can use these n replicates as blocks, and assign the
treatments to the experimental units within each of the n blocks. If we are going to replicate the
experiment anyway, at almost no additional cost, you can block the experiment, doing one replicate
first, then the second replicate, etc. rather than completely randomize the n times treatment
combinations to all the runs.
There is almost always an advantage to blocking when we replicate the treatments. This is true even
if we only block using time due to the order of the replicates. However, there are often many other
factors that we have available as potential sources of variation that we can include as a block factor,
such as batches of material, technician, day of the week, or time of day, or other environmental
factors. Thus if we can afford to replicate the design then it is almost always useful to block.
As we have disclosed in chapter 4, blocking is advantageous in noise reduction. Suppose that the
factorial design has been replicated n times. The analysis of factorial design with block is
the same as what we have seen so far in this chapter. To have a better understanding, let us
consider the following design with r blocks and see how the sum square for block is computed
157
Confounding is a design technique for arranging a complete factorial experiment in blocks,

where the block size is smaller than the number of treatment combinations in one replicate. The
technique causes information about certain treatment effects (usually high order interaction) to
be indistinguishable from, or confounded with, blocks. We will consider the construction of
and designs in two incomplete blocks.
Suppose we wish to run a single replicate of and designs. From our discussion of
design, we can estimate the main effects using the usual formula; this technique is not affected
whether blocking is there or not. That is,
Let us consider A, and a have plus sign and put in one block and and (1) have minus sign
and put them in the other block. Note that each block contains two treatment combinations in
fact we have four treatment combinations. That is why we call it the block incomplete. In the
above case we can say that A is confounded with block.
158
When we confound B, ab and b have plus sign and put in one block and a
and (1) have minus sign and put them in the other block.
Now consider the higher-order, AB interaction
Because the two treatment combinations with the plus sign ( and (1)) are in one block and the
two with minus sign ( and ) are in block 2, the block effect and interaction are identical.
That is, is confounded with block.
Above we have seen the possibilities of confounding A, B and AB. The usual practice is to
confound the highest - order interaction with blocks. That is, AB and ABC in case of
designs, respectively.
Note that: when the number of variables is small, say 𝐾 𝑜𝑟 , it is usually necessary to replicate the
experiment to obtain an estimate of error.
159

Chapter 5 Design of Experiment

Uploaded by

Copyright:

Available Formats

Chapter 5 Design of Experiment

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 5 Design of Experiment

Uploaded by

Copyright:

Available Formats

Lecture notes for Design and analysis of experiments (Stat 2043) Chapter - 5

5.1 Basic definitions and principles

5.2 The advantage of factorial designs

5.3 The two factor factorial design

5.3.1 Two factor design without interaction

5.3.2 Two factor design with interaction

5.4 The three factor factorial design

5.5 Blocking in a factorial design

5.6 Unbalanced data in a factorial design

5.7 The 2k factorial design

5.7.1 The 22 design

5.7.2 The 23 design

5.8 Blocking and confounding in 22 and 23 designs

considered what they do not say”. W. WATT

5.1 Basic definitions and principles

5.2 The advantage of factorial designs

As we know in chapter one, factorial experiment is one of the strategies of experimentation. A

The Advantage of Factorial designs

Factorial designs have several advantages some of them are:

5.3 The two factor factorial design

Then the layout of data looks like the following:

and the equality of column treatment effects, say

5.3.1 Two factor design without interaction

The total corrected sum of squares can be rewritten as

 For a sake of simplifying computation of sum of squares

Source of variation Sum of squares DF Mean square

The corresponding degrees of freedom are

o The appropriate mode is then

5.3.2 Two factor design with interaction

For simplicity of computation,

We may also compute by subtraction as follows:

The ANOVA table looks like;

Source of variation Sum of squares DF Mean square

NB: DF: Degree of Freedom

Step 1: State the null hypothesis and alternative hypothesis.

i, for factor A (for Material type)

ii, for factor B ( for Temperature)

iii, for interaction ( AB (Material*Temperature) )

Step 3: Identify the test statistic, F – test.

Step 4: Determine test statistic.

Step 5: Identify the critical value

= = = 3.35,for Material type

Step 6: Make decision:

o Reject if , , for Material type.

o Reject if , ) for Temperature.

o Reject if , for interaction.

 3.56 > 2.73 so reject for interaction (Material*Temperature).

At 5% level of significance we have enough evidence to conclude that

there is significance difference among Material type.

At 5% level of significance we have enough evidence to conclude that

there is significance difference among Temperature.

At 5% level of significance we have enough evidence to conclude that

there is significance difference among interaction (At 5% level of

significance we have enough evidence to conclude that there is a

significant interaction between material type and temperature).

The ANOVA table becomes:

Source of variation Sum of squares DF Mean square

Material Type a-1

Source of variation Sum of squares DF Mean square

Material Type 10683.72 2 5341.86 7.91

Solution: , abn = 324 = 24, = 2278(the sum of all observations)