Design and

Analysis of
Experiments
Chapter 6: The 2
k
Factorial
Design
Introduction
Special case of the general factorial design; k factors, all
at two levels
A complete replicate of such a design requires 2 2
2 = 2
k
observations and is called a 2
k
Factorial Design
The two levels are usually called low and high (they could
be either quantitative or qualitative)
Very widely used in industrial experimentation (factor
screening experiments)
we assume that
(1) the factors are fixed,
(2) the designs are completely randomized,
(3) the usual normality assumptions are satisfied.
(4) the response is approximately linear
The Simplest Case: The 2
2
- and + denote the low and
high levels of a factor,
respectively
Low and high are arbitrary
terms
Geometrically, the four runs
form the corners of a
square
Factors can be quantitative
or qualitative, although their
treatment in the final model
will be different
Analysis Procedure for a
Factorial Design
Estimate factor effects
Formulate model
With replication, use full model
With an unreplicated design, use normal probability
plots
Statistical testing (ANOVA)
Refine the model
Analyze residuals (graphical)
Interpret results
Estimation of Factor Effects
a :A at the high level and B
at the low level
b : A at the low level and B
at the high level,
ab represents both factors
at the high level
(1) both factors at the low
level.
1
2
1
2
1
2
(1)
2 2
[ (1)]
(1)
2 2
[ (1)]
(1)
2 2
[ (1) ]
A A
n
B B
n
n
A y y
ab a b
n n
ab a b
B y y
ab b a
n n
ab b a
ab a b
AB
n n
ab a b
+
+
=
+ +
=
= +
=
+ +
=
= +
+ +
=
= +
The effect estimates are: A
= 8.33, B = -5.00, AB = 1.67
Statistical Testing - ANOVA
Examine the magnitude and direction of the
factor effects
ANOVA can generally be used to confirm this
interpretation.
Total effects of A, B and AB are estimated in the
form of contrasts for example:
ContrastA = ab + a - b - (1)
In the standard order:
Based on the P-values, we conclude that the main effects
are statistically significant and that there is no interaction
between these factors.
The Regression Model
It is easy to express the results of the experiment in terms
of a regression model.

Where X1 and X2 are coded variables for reactant
concentration and amount of catalyst respectively.
The relationship between the natural variables and coded
variables:



The fitted regression model is:
The Response Surface

The regression model can be used to generate
response surface plots.
Substituting the coded variables with the normal
variables we have:





we use a fitted surface such as this to find a
direction of potential improvement for a process.
Because the model is first-order (that is, it contains
only the main effects), the fitted response surface
is a plane.
Residuals and Model Adequacy
The residuals are the differences between the
observed and fitted values of y. For example,
when the reactant concentration is at the low
level (X1 = -1) and the catalyst is at the low level
(X2 = -1), the predicted yield is:
Residuals and Model Adequacy
The 2
3
Factorial Design
three factors, A, B, and
C, each at two levels
and the eight treatment
combinations can be
displayed geometrically
as a cube.
Effects in The 2
3
Factorial Design

Properties of the Table
Except for column I, every column has an equal number of
+ and signs
The sum of the product of signs in any two columns is zero
Multiplying any column by I leaves that column unchanged
(identity element)
The product of any two columns yields a column in the
table:




Orthogonal design
Orthogonality is an important property shared by all
factorial designs
2
A B AB
AB BC AB C AC
=
= =
Statistical analysis (ANOVA)
Sums of squares for the effects are easily computed. In
the 2
3
design with n replicates, the sum of squares for
any effect is:

The Regression Model and Response Surface
Unreplicated 2
k
Factorial Designs
2
k
factorial designs with one observation for
each experimental run
An unreplicated 2
k
factorial design is also
sometimes called a single replicate
These designs are very widely used
Lack of replication causes potential problems in
statistical testing
Replication admits an estimate of pure error
With no replication, fitting the full model results in zero
degrees of freedom for error (Modeling the noise)
chance of unusual response observations
Spacing of Factor Levels in the
Unreplicated 2
k
Factorial Designs

If the factors are spaced too closely, it increases the chances
that the noise will overwhelm the signal in the data
More aggressive spacing is usually best
solutions to this problem
Sparsity of effects principle: Pooling high-order
interactions to estimate error.

Normal probability plotting of effects (Daniels, 1959)
The effects that are negligible are normally distributed, with mean
zero and variance and will tend to fall along a straight line on this
plot
significant effects will have nonzero means and will not lie along
the straight line. Thus the preliminary model will be
Half-normal plot
This is a plot of the absolute value of the effect estimates against
their cumulative normal probabilities.
A Single Replicate of the 2
4
Design
Investigation the effects of four factors on the
filtration rate of a resin
The factors are A = temperature, B = pressure, C = mole
ratio, D= stirring rate



Estimates of the Effects
The Half-Normal Probability Plot of
Effects

Design Projection: ANOVA Summary for the
Model as a 2
3
in Factors A, C, and D

The Regression Model
The coded variables x1, x3, x4 take on values between -
1 and + 1. The predicted filtration rate at run (1) is
Model Residuals are Satisfactory

Model Interpretation Main Effects and
Interactions
Model Interpretation Response
Surface Plots
With concentration at either the low or high level, high
temperature and high stirring rate results in high filtration rates
Other Methods for Analyzing
Unreplicated Factorials
Lenth's method:The basis of Lenth's method is
to estimate the variance of a contrast from the
smallest (in absolute value) contrast estimates.
"pseudo standard error"
margin of error
simultaneous margin of error
For Filtration Example
Adjusted multipliers for Lenths method
the original method makes too many type I errors,
especially for small designs (few contrasts)
To address the problem adjusted multipliers have been
suggested:

Lenths method is a nice supplement to the normal
probability plot of effects
Conditional inference chart
The purpose of the graph is to help the experimenter in
judging significant effects.
In unreplicated designs, there is no internal estimate of
variance
conditional inference chart is designed to help the
experimenter evaluate effect magnitude for a range of
standard deviation values.

The Drilling Experiment:Data
Transformation in a Factorial Design

A = drill load, B = flow, C = speed, D = type of mud,
y = advance rate of the drill
Normal Probability Plot of Effects
The Drilling Experiment

Residual Plots

The residual plots indicate that there are problems with
the equality of variance assumption
The usual approach to this problem is to employ a
transformation on the response
Power family transformations are widely used


Transformations are typically performed to
Stabilize variance
Induce at least approximate normality
Simplify the model
*
y y

=
Effect Estimates Following the
Log Transformation
Three main effects are
large
No indication of large
interaction effects
What happened to the
interactions?
expressing the data in the
correct metric has
simplified its structure to
the point that the two
interactions are no longer
required in the
explanatory model.
Residuals

Location and Dispersion Effects in an
Unreplicated Factorial



Duplicate Measurements on the
Response

The four design factors are temperature (A), time
(B), pressure (C), and gas flow (D).




Addition of Center Points
to a 2
k
Designs
Runs at the center provide an estimate of
error and allow the experimenter to
distinguish between two possible models:
0
1 1
2
0
1 1 1
First-order model (interaction)
Second-order model
k k k
i i ij i j
i i j i
k k k k
i i ij i j ii i
i i j i i
y x x x
y x x x x
| | | c
| | | | c
= = >
= = > =
= + + +
= + + + +



no "curvature"
F C
y y =
The hypotheses are:
0
1
1
1
: 0
: 0
k
ii
i
k
ii
i
H
H
|
|
=
=
=
=

2
Pure Quad
( )
F C F C
F C
n n y y
SS
n n

=
+
This sum of squares has a
single degree of freedom
Central Composite Design