Design of Engineering Experiments Part 5 - The 2 Factorial Design

Design of Engineering Experiments
Part 5 – The 2k Factorial Design

• Text reference, Chapter 6
• Special case of the general factorial design; k factors, all
at two levels
• The two levels are usually called low and high (they
could be either quantitative or qualitative)
• Very widely used in industrial experimentation
• Form a basic “building block” for other very useful
experimental designs (DNA)
• Special (short-cut) methods for analysis
• We will make use of Design-Expert
Chapter 6 Design & Analysis of Experiments 1

7E 2009 Montgomery
7E 2009 Montgomery
The Simplest Case: The 22
“-” 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

7E 2009 Montgomery
Chemical Process Example
A = reactant concentration, B = catalyst amount,

y = recovery

7E 2009 Montgomery
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

7E 2009 Montgomery
Estimation of Factor Effects
A  y A   y A
See textbook, pg. 209-210 For
ab  a b  (1)
  manual calculations
2n 2n
 21n [ab  a  b  (1)] The effect estimates are:
A = 8.33, B = -5.00, AB = 1.67
B  yB  yB
ab  b a  (1) Practical interpretation?
 
2n 2n Design-Expert analysis
 21n [ab  b  a  (1)]
ab  (1) a  b
AB  
2n 2n
 21n [ab  (1)  a  b]
7E 2009 Montgomery
Form Tentative Model
Term Effect SumSqr % Contribution
Model Intercept
Model A 8.33333 208.333 64.4995
Model B -5 75 23.2198
Model AB 1.66667 8.33333 2.57998
Error Lack Of Fit 0 0
Error P Error 31.3333 9.70072
Lenth's ME 6.15809
Lenth's SME 7.95671

7E 2009 Montgomery
Statistical Testing - ANOVA
The F-test for the “model” source is testing the significance of the
overall model; that is, is either A, B, or AB or some combination of
these effects important?

7E 2009 Montgomery
Design-Expert output, full model

7E 2009 Montgomery
Design-Expert output, edited
or reduced model

7E 2009 Montgomery
Residuals and Diagnostic Checking

7E 2009 Montgomery
The Response Surface

7E 2009 Montgomery
The 23 Factorial Design

7E 2009 Montgomery
Effects in The 23 Factorial Design
A  y A  y A
B  yB  y B
C  yC   yC 
etc, etc, ...
Analysis
done via
computer

7E 2009 Montgomery
An Example of a 23 Factorial Design
A = gap, B = Flow, C = Power, y = Etch Rate

7E 2009 Montgomery
Table of – and + Signs for the 23 Factorial Design (pg. 218)

7E 2009 Montgomery
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:
A  B  AB
AB  BC  AB 2C  AC
• Orthogonal design
• Orthogonality is an important property shared by all factorial designs

7E 2009 Montgomery

7E 2009 Montgomery
ANOVA Summary – Full Model

7E 2009 Montgomery
Model Coefficients – Full Model

7E 2009 Montgomery
Refine Model – Remove Nonsignificant Factors

7E 2009 Montgomery
Model Coefficients – Reduced Model

7E 2009 Montgomery
Model Summary Statistics for Reduced Model
• R2 and adjusted R2
SS Model 5.106 10 5

R  2
  0.9608
SST 5.314 10 5
SS E / df E 20857.75 /12
R 2
Adj  1  1  0.9509
SST / dfT 5.314 10 /15
5
• R2 for prediction (based on PRESS)

PRESS 37080.44
R 2
Pred  1  1  0.9302
SST 5.314 10 5

7E 2009 Montgomery
Model Summary Statistics
• Standard error of model coefficients (full

model)
ˆ ˆ  2
MS E 2252.56
se(  )  V (  )  k
 k
  11.87
n2 n2 2(8)
• Confidence interval on model coefficients
ˆ  t / 2,df E se( ˆ )    ˆ  t / 2,dfE se( ˆ )

7E 2009 Montgomery
The Regression Model

7E 2009 Montgomery
Model Interpretation
Cube plots are

often useful visual
displays of
experimental
results

7E 2009 Montgomery
Cube Plot of Ranges
What do the
large ranges
when gap and
power are at the
high level tell
you?

7E 2009 Montgomery
7E 2009 Montgomery
The General 2k Factorial Design
• Section 6-4, pg. 227, Table 6-9, pg. 228
• There will be k main effects, and
k 
  two-factor interactions
 2
k 
  three-factor interactions
 3

1 k  factor interaction
7E 2009 Montgomery
6.5 Unreplicated 2k Factorial Designs
• These are 2k factorial designs with one
observation at each corner of the “cube”
• An unreplicated 2k factorial design is also
sometimes called a “single replicate” of the 2k
• These designs are very widely used
• Risks…if there is only one observation at each
corner, is there a chance of unusual response
observations spoiling the results?
• Modeling “noise”?

7E 2009 Montgomery
Spacing of Factor Levels in the
Unreplicated 2k 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
7E 2009 Montgomery
Unreplicated 2k Factorial Designs
• Lack of replication causes potential problems in
statistical testing
– Replication admits an estimate of “pure error” (a better
phrase is an internal estimate of error)
– With no replication, fitting the full model results in
zero degrees of freedom for error
• Potential solutions to this problem
– Pooling high-order interactions to estimate error
– Normal probability plotting of effects (Daniels, 1959)
– Other methods…see text

7E 2009 Montgomery
Example of an Unreplicated 2k Design
• A 24 factorial was used to investigate 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
• Experiment was performed in a pilot plant

7E 2009 Montgomery
The Resin Plant Experiment

7E 2009 Montgomery
The Resin Plant Experiment

7E 2009 Montgomery
7E 2009 Montgomery
Estimates of the Effects

7E 2009 Montgomery
The Half-Normal Probability Plot of Effects

7E 2009 Montgomery
Design Projection: ANOVA Summary for
the Model as a 23 in Factors A, C, and D

7E 2009 Montgomery
The Regression Model

7E 2009 Montgomery
Model Residuals are Satisfactory

7E 2009 Montgomery
Model Interpretation – Main Effects
and Interactions

7E 2009 Montgomery
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
7E 2009 Montgomery
Outliers: suppose that cd = 375 (instead of 75)

7E 2009 Montgomery
Dealing with Outliers
• Replace with an estimate
• Make the highest-order interaction zero
• In this case, estimate cd such that ABCD =
0
• Analyze only the data you have
• Now the design isn’t orthogonal
• Consequences?
7E 2009 Montgomery
7E 2009 Montgomery
The Drilling Experiment
Example 6.3
A = drill load, B = flow, C = speed, D = type of mud,

y = advance rate of the drill

7E 2009 Montgomery
Normal Probability Plot of Effects –
The Drilling Experiment

7E 2009 Montgomery
Residual Plots
Residuals vs. Predicted

2. 58625
1. 44875
R es id uals
0. 31125
- 0. 82625
- 1. 96375
1.6 9 4. 70 7. 70 10. 71 13. 71
Pre d i c te d

7E 2009 Montgomery
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

y y
*
• Transformations are typically performed to
– Stabilize variance
– Induce at least approximate normality
– Simplify the model

7E 2009 Montgomery
Selecting a Transformation
• Empirical selection of lambda
• Prior (theoretical) knowledge or experience can
often suggest the form of a transformation
• Analytical selection of lambda…the Box-Cox
(1964) method (simultaneously estimates the
model parameters and the transformation
parameter lambda)
• Box-Cox method implemented in Design-Expert

7E 2009 Montgomery
(15.1)

7E 2009 Montgomery
DESIG N-EXPERT Plo t
ad v._ra te
La mb da
Curre nt = 1
Best = -0 .23
Lo w C. I. = -0 .7 9
High C.I. = 0 .3 2
Recom m en d tran sf orm :

Lo g
(L am bd a = 0 )
The Box-Cox Method
Box-Cox Plot for Power Transforms
6.85 A log transformation is
recommended
5.40 The procedure provides a
confidence interval on
Ln(Res idualSS)
the transformation
3.95
parameter lambda
If unity is included in the
2.50 confidence interval, no
transformation would be
needed
1.05
-3 -2 -1 0 1 2 3
Lam bda

7E 2009 Montgomery
Effect Estimates Following the
Log Transformation
Three main effects are

large
No indication of large
interaction effects
What happened to the
interactions?

7E 2009 Montgomery
ANOVA Following the Log Transformation

7E 2009 Montgomery
Following the Log Transformation

7E 2009 Montgomery
The Log Advance Rate Model
• Is the log model “better”?
• We would generally prefer a simpler model
in a transformed scale to a more
complicated model in the original metric
• What happened to the interactions?
• Sometimes transformations provide insight
into the underlying mechanism

7E 2009 Montgomery
Other Examples of
Unreplicated 2k Designs
• The sidewall panel experiment (Example 6.4, pg. 245)
– Two factors affect the mean number of defects
– A third factor affects variability
– Residual plots were useful in identifying the dispersion
effect
• The oxidation furnace experiment (Example 6.5, pg.
245)
– Replicates versus repeat (or duplicate) observations?
– Modeling within-run variability
7E 2009 Montgomery
Other Analysis Methods for
Unreplicated 2k Designs
• Lenth’s method (see text, pg. 235)
– Analytical method for testing effects, uses an estimate
of error formed by pooling small contrasts
– Some adjustment to the critical values in the original
method can be helpful
– Probably most useful as a supplement to the normal
probability plot
• Conditional inference charts (pg. 236)

7E 2009 Montgomery
Overview of Lenth’s method
For an individual contrast, compare to the margin of error

7E 2009 Montgomery
7E 2009 Montgomery
Adjusted multipliers for Lenth’s method
Suggested because the original method makes too many
type I errors, especially for small designs (few contrasts)
Simulation was used to find these adjusted multipliers

Lenth’s method is a nice supplement to the normal
probability plot of effects
JMP has an excellent implementation of Lenth’s method
in the screening platform
7E 2009 Montgomery
7E 2009 Montgomery
The 2k design and design optimality
The model parameter estimates in a 2k design (and the effect estimates) are
least squares estimates. For example, for a 22 design the model is
y   0  1 x1   2 x2  12 x1 x2  
(1)   0  1 (1)   2 (1)  12 ( 1)(1)  1

a   0  1 (1)   2 (1)  12 (1)(1)   2 The four
b   0  1 (1)   2 (1)  12 (1)(1)   3 observations
from a 22 design
ab   0  1 (1)   2 (1)  12 (1)(1)   4
 (1)  1 1 1 1   0   1 
a 1 1 1 1    
y = Xβ + ε, y    , X    ,β   1  ,ε   2 
b 1 1 1 1  2   3 
       
ab
  1 1 1 1  
 12   4 
7E 2009 Montgomery
The least squares estimate of β is
βˆ = (XX)-1 Xy The “usual” contrasts

1
4 0 0 0 (1)  a  b  ab 
0 4 0 0  a  ab  b  (1)  The XX matrix is
   diagonal –
0 0 4 0 b  ab  a  (1)  consequences of an
  
0 0 0 4  (1)  a  b  ab  orthogonal design
 (1)  a  b  ab 
 4 
 ˆ0  (1)  a  b  ab   a  ab  b  (1) 
     
 ˆ1  1  a  ab  b  (1)   4 
 ˆ   I 4 b  ab  a  (1)    b  ab  a  (1)  The regression
 2  4     coefficient estimates
ˆ   (1)  a  b  ab   4  are exactly half of the

 12   (1)  a  b  ab  ‘usual” effect estimates
 
 4 
7E 2009 Montgomery
The matrix XX has interesting and useful properties:
V ( ˆ )   2 (diagonal element of (XX) 1 )

2 Minimum possible
 value for a four-run
4 design
Maximum possible
|(XX) | 256 value for a four-run
design
Notice that these results depend on both the design that you
have chosen and the model
What about predicting the response?

7E 2009 Montgomery
V [ yˆ ( x1 , x2 )]   2 x(XX)-1 x
x  [1, x1 , x2 , x1 x2 ]
2
V [ yˆ ( x1 , x2 )]  (1  x12  x22  x12 x22 )
4
The maximum prediction variance occurs when x1  1, x2  1
V [ yˆ ( x1 , x2 )]   2
The prediction variance when x1  x2  0 is
2
V [ yˆ ( x1 , x2 )] 
4
What about average prediction variance over the design space?

7E 2009 Montgomery
Average prediction variance
1 1
1
I    V [ yˆ ( x1 , x2 )dx1dx2 A = area of design space = 2 2  4
A 1 1
1 1
1 2 1
    (1  x12  x22  x12 x22 )dx1dx2
4 1 1 4
4 2

9

7E 2009 Montgomery
D e si g n -E xp e rt® S o f tw a r e
M i n StdEr r Me an: 0. 500

M a x S td E r r M e a n : 1 . 0 0 0
C u boi da l
r adi us = 1
P o in ts = 1 0 0 0 0
FDS Graph
1.000
0.750
StdErr Mean 0.500
0.250
0.000
0.00 0.25 0.50 0.75 1.00
Fraction of Design Space

7E 2009 Montgomery
For the 22 and in general the 2k
• The design produces regression model coefficients that
have the smallest variances (D-optimal design)
• The design results in minimizing the maximum
variance of the predicted response over the design space
(G-optimal design)
• The design results in minimizing the average variance
of the predicted response over the design space (I-
optimal design)

7E 2009 Montgomery
Optimal Designs
• These results give us some assurance that
these designs are “good” designs in some
general ways
• Factorial designs typically share some (most)
of these properties
• There are excellent computer routines for
finding optimal designs (JMP is outstanding)

7E 2009 Montgomery
Addition of Center Points
to a 2k Designs
• Based on the idea of replicating some of the
runs in a factorial design
• Runs at the center provide an estimate of
error and allow the experimenter to
distinguish between two possible models:
k k k
First-order model (interaction) y   0    i xi    ij xi x j  
i 1 i 1 j i
k k k k
Second-order model y   0    i xi    ij xi x j    ii xi2  
i 1 i 1 j i i 1

7E 2009 Montgomery
7E 2009 Montgomery
yF  yC  no "curvature"
The hypotheses are:
k
H 0 :   ii  0
i 1
k
H1 :   ii  0
i 1
nF nC ( yF  yC ) 2
SS Pure Quad 
nF  nC
This sum of squares has a

single degree of freedom

7E 2009 Montgomery
Example 6.6, Pg. 248
Refer to the original experiment
shown in Table 6.10. Suppose that nC  4
four center points are added to this
experiment, and at the points x1=x2 Usually between 3
=x3=x4=0 the four observed and 6 center points
filtration rates were 73, 75, 66, and will work well
69. The average of these four
Design-Expert
center points is 70.75, and the
provides the analysis,
average of the 16 factorial runs is
including the F-test
70.06. Since are very similar, we
for pure quadratic
suspect that there is no strong
curvature
curvature present.

7E 2009 Montgomery
7E 2009 Montgomery
ANOVA for Example 6.6 (A Portion of Table 6.22)

7E 2009 Montgomery
If curvature is significant, augment the design with axial runs to
create a central composite design. The CCD is a very effective design
for fitting a second-order response surface model

7E 2009 Montgomery
Practical Use of Center Points (pg. 260)
• Use current operating conditions as the center
point
• Check for “abnormal” conditions during the
time the experiment was conducted
• Check for time trends
• Use center points as the first few runs when there
is little or no information available about the
magnitude of error
• Center points and qualitative factors?

7E 2009 Montgomery
Center Points and Qualitative Factors

7E 2009 Montgomery

Design of Engineering Experiments Part 5 - The 2 Factorial Design

Uploaded by

Copyright:

Available Formats

You might also like

Design of Engineering Experiments Part 5 - The 2 Factorial Design

Uploaded by

Document Information

Original Title

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

Design of Engineering Experiments Part 5 - The 2 Factorial Design

Uploaded by

Copyright:

Available Formats

Design of Engineering Experiments

Part 5 – The 2k Factorial Design

Chapter 6 Design & Analysis of Experiments 1

“-” and “+” denote the low and

Chapter 6 Design & Analysis of Experiments 3

A = reactant concentration, B = catalyst amount,

Chapter 6 Design & Analysis of Experiments 4

Chapter 6 Design & Analysis of Experiments 5

Chapter 6 Design & Analysis of Experiments 7

Chapter 6 Design & Analysis of Experiments 8

Chapter 6 Design & Analysis of Experiments 9

Chapter 6 Design & Analysis of Experiments 10

Chapter 6 Design & Analysis of Experiments 11

Chapter 6 Design & Analysis of Experiments 12

Chapter 6 Design & Analysis of Experiments 13

Chapter 6 Design & Analysis of Experiments 14

A = gap, B = Flow, C = Power, y = Etch Rate

Chapter 6 Design & Analysis of Experiments 15

Chapter 6 Design & Analysis of Experiments 16

Chapter 6 Design & Analysis of Experiments 17

Chapter 6 Design & Analysis of Experiments 18

Chapter 6 Design & Analysis of Experiments 19

Chapter 6 Design & Analysis of Experiments 20

Chapter 6 Design & Analysis of Experiments 21

Chapter 6 Design & Analysis of Experiments 22

SS Model 5.106 10 5

• R2 for prediction (based on PRESS)

Chapter 6 Design & Analysis of Experiments 23

• Standard error of model coefficients (full

• Confidence interval on model coefficients

ˆ  t / 2,df E se( ˆ )    ˆ  t / 2,dfE se( ˆ )

Chapter 6 Design & Analysis of Experiments 24

Chapter 6 Design & Analysis of Experiments 25

Cube plots are

Chapter 6 Design & Analysis of Experiments 26

Chapter 6 Design & Analysis of Experiments 27

Chapter 6 Design & Analysis of Experiments 30

If the factors are spaced too closely, it increases the chances

Chapter 6 Design & Analysis of Experiments 32

• A 24 factorial was used to investigate the

Chapter 6 Design & Analysis of Experiments 33

Chapter 6 Design & Analysis of Experiments 34

Chapter 6 Design & Analysis of Experiments 35

Chapter 6 Design & Analysis of Experiments 37

Chapter 6 Design & Analysis of Experiments 38

Chapter 6 Design & Analysis of Experiments 39

Chapter 6 Design & Analysis of Experiments 40

Chapter 6 Design & Analysis of Experiments 41

Chapter 6 Design & Analysis of Experiments 42

Chapter 6 Design & Analysis of Experiments 44

A = drill load, B = flow, C = speed, D = type of mud,

Chapter 6 Design & Analysis of Experiments 47

Chapter 6 Design & Analysis of Experiments 48

Residuals vs. Predicted

1.6 9 4. 70 7. 70 10. 71 13. 71

Chapter 6 Design & Analysis of Experiments 49

Chapter 6 Design & Analysis of Experiments 50

Chapter 6 Design & Analysis of Experiments 51