Anova - Example - P60 - 69

What If There Are More Than
Two Factor Levels?

• Chapter 3
• Comparing more that two factor levels…the analysis of

variance
• ANOVA decomposition of total variability
• Statistical testing & analysis
• Checking assumptions, model validity
• Post-ANOVA testing of means
ANOVA_EXAMPLE P60 Brainerd 1

What If There Are More Than
Two Factor Levels?
• The t-test does not directly apply
• There are lots of practical situations where there are either
more than two levels of interest, or there are several factors
of simultaneous interest
• The analysis of variance (ANOVA) is the appropriate
analysis “engine” for these types of experiments – Chapter
3, textbook
• The ANOVA was developed by Fisher in the early 1920s,
and initially applied to agricultural experiments
• Used extensively today for industrial experiments
An Example (See pg. 60)
• Consider an investigation into the formulation of a new
“synthetic” fiber that will be used to make cloth for shirts
• The response variable is tensile strength
• The experimenter wants to determine the “best” level of
cotton (in wt %) to combine with the synthetics
• Cotton content can vary between 10 – 40 wt %; some non-
linearity in the response is anticipated
• The experimenter chooses 5 levels of cotton “content”;
15, 20, 25, 30, and 35 wt %
• The experiment is replicated 5 times – runs made in
random order

ANOVA: Design of Experiments
Chapter 3
•A
A
A product
productdevelopment
product development
development engineer
engineer isisinterested
engineer interested inininvestigating
is interested investigating the
thetensile
in investigatingtensile
strength
the tensile of a strength
new syntheticof afiber
newthat will be used
synthetic fiber to that
makewillmen’s beshirts.
used The
to
engineer knows from previous experience that the strength is affected by the
make
•weight
men’s
The engineer
shirts.
percent knows
of cotton
The
from
used
engineer
previous
in the blend
knows
experience
from previous
of materials
that theforstrength
the fiber.
is affected
experience
Furthermore,
by the weight hethat theofstrength
percent
suspects cotton is affected
that increasing
used in the
theblend byofthe
cotton weight
content
materials will percent
forincrease
the fiber.of
the
cotton
strength,used ininitially.
at least the blend Heof materials
also knows thatfor the fiber.
cotton content Furthermore,
should range
•between
heFurthermore,
aboutthat
suspects 10
he suspects
percent and
that40
increasing increasing
percent
the cottonifthe
the cotton
final cloth
content content
willis increase
to
will
haveincrease
other
the
quality
the strength,
strength,characteristics
atatleast
least initially.
initially.
that are He
desired.
also knows
The engineer
that cotton decides
contentto test
should
specimens
range between
at five
about
levels
10 of
percent
cottonandweight
40 percent
percent: if the
15 percent,
final cloth20 ispercent,
to have25
percent,
other quality
30 percent,
characteristics
and 35 percent.
that are desired.
•The engineer decides to test specimens at five levels of cotton weight

percent: 15 percent, 20 percent, 25 percent, 30 percent, and 35 percent. He
also decides to test five specimens at each level of cotton content.

Chapter 3
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
H a : At least one µ i is different
Tensile
Strength
EXAMPLE PROBLEM
A single-factor experiment with a = 5 levels
of the factor and n = 5 replicates. The 25
runs should be made in random order.
15 20 25 30 35 Cotton %

Chapter 3
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
Tensile
Strength ONE FROM EACH
15 20 25 30 35 Cotton %
Can we determine which is better?
Chapter 3
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
REPLICATION
Tensile
Strength
15 20 25 30 35 Cotton %
What does replication provide?
Chapter 3
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
REPLICATION
Tensile Effect
Strength If the sample mean is used
to estimate the effect of a
factor in the experiment,
then replication permits
the experimenter to obtain
a more precise estimate of
this effect.
15 20 25 30 35 Cotton %
What else does replication provide?
Chapter 3
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
REPLICATION
Tensile Effect
Strength Error
Allows the experimenter to
σ 2
obtain an estimate of the
experimental error. This
σ 2
=
estimate if error becomes a
xi of measurement
basic unit
n
for determining whether
observed differences in the
data are really statistically
15 20 25 30 35 different. Cotton %
What assumption does the error estimate depend upon?
Chapter 3
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
REPLICATION
Tensile Effect
Strength Error
RANDOMIZATION
Both the allocation of the
experimental material and the
order in which the individual
runs or trials of the experiment
are to be performed are
randomly determined.
Cotton % 15 20 25 30 35

Chapter 3
Need to randomize run order
COTTON% Experimental Run #

15 1 2 3 4 5
20 6 7 8 9 10
25 11 12 13 14 15
30 16 17 18 19 20
35 21 22 23 24 25

Chapter 3
Test Seq Run # %
1 8 20
2 18 30
3 10 20
4 23 35
5 17 30
* * *
25 3 15
RANDOMIZE RUNS
Chapter 3
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
REPLICATION
Tensile Effect
Strength Error
RANDOMIZATION
Cotton %
15 20 25 30 35
What if the measurements varied widely because of human operators?
Chapter 3
• PUSHUP EXAMPLE
– Test if one can do push ups better in the
morning or afternoon. 20 DATA POINTS
– Select 40 people at random
Is PM really
better than AM?
AM PM
Chapter 3
• What if the PM group of 20 was in better shape then the AM group of
20?
• What if the test was conducted on a Monday morning?
• What if different people counted the push ups between AM and PM?
Is PM really
better than AM?
AM PM

Chapter 3
Controllable Factors
Input PROCESS Output
Uncontrollable Factors

Chapter 3
• BLOCKING
– Used to limit the uncontrollable factors
– Therefore increase precision
• PUSH UP EXAMPLE
– Paired Data = BLOCKING
– Have same person do AM and PM
– You are investigating AM vs PM not which group can
do more pushups.
– Randomly Sort experiment by Days of the Week and
have one grader

THE THREE PRINCIPLES of Experimental
Design
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
Tensile
REPLICATION
Strength
RANDOMIZATION
BLOCKING
A block is a portion of the
experimental material that
should be more homogeneous
than the entire set of material.
15 20 25 30 35 Cotton %

Analysis of Variance (ANOVA)
H 0 : µ1 = µ 2 = µ 3 = ... = µ i Variation
Between
H a : At least one µ i is different Samples
Tensile
Strength
Variation
Within 15 20 25 30 35 Cotton %
Samples
When very different
Tensile
Strength
Between Sample
Variation Large
Within Sample
Variation Small
15 20 25 30 35 Cotton %

When near equal
Tensile
Strength
15 20 25 30 35 Cotton %
Between Sample Variation near equal to
Within Sample Variation
Between Sample Variation
TEST STAT =
F-TEST
REQUIRED ASSUMPTION
All data is normal with equal variance
EXTENSION OF TWO SAMPLE POOLED t

SETUP (i = factor; j = replicate)
Level Replicates Mean SD

1 2 3 * j
1 11 12 13 * 1j 1*
2 21 22 23 * 2j 2*
3 31 32 33 * 3j 3*
* * * * * *
i i1 i2 i3 * ij
**
xi , j = observed value i th level, jth measurement
COTTON EXAMPLE
Level Replicates Mean SD

1 2 3 4 5
1 7 7 15 11 9 9.80 3.35
2 12 17 12 18 18 15.40 3.13
3 14 18 18 19 19 17.60 2.07
4 19 25 22 19 23 21.60 2.61
5 7 10 11 15 11 10.80 2.86
15.04
xi• = sample mean i th level
x•• = Grand Mean
I = # factors and J = # replicates
Mean Square Treatment ≡ Between Sample Variation

J
MSTr = ∑
I −1 i
( X i • − X •• ) 2
Mean Square Error ≡ Within Sample Variation
S + S + S + ....S
2 2 2 2
MSE = 1 2 3 I
I
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
Between Sample Variation
TEST STAT =
MSTr If H 0 true
f = E ( MSTr ) = E ( MSE ) = σ 2
MSE
If H 0 false
Fα , I −1, I ( J −1) E ( MSTr ) > E ( MSE ) = σ 2

Anova Chapter 3 (See pg. 62)
• Does changing the

cotton weight percent
change the mean
tensile strength?
• Is there an optimum
level for cotton
content?
The Analysis of Variance
SST = SSTreatments + SS E
• A large value of SSTreatments reflects large differences in
treatment means
• A small value of SSTreatments likely indicates no differences in
treatment means
• Formal statistical hypotheses are:
H 0 : µ1 = µ 2 = L = µ a
H1 : At least one mean is different

The Analysis of Variance
• While sums of squares cannot be directly compared to test
the hypothesis of equal means, mean squares can be
compared. ( MS = Estimates of variances)
• A mean square is a sum of squares divided by its degrees
of freedom:
dfTotal = dfTreatments + df Error
an − 1 = a − 1 + a ( n − 1)
SSTreatments SS E
MSTreatments = , MS E =
a −1 a ( n − 1)
• If the treatment means are equal, the treatment and error
mean squares will be (theoretically) equal.
• If treatment means differ, the treatment mean square will
be larger than the error mean square.
The Analysis of Variance is
Summarized in a Table
• Computing…see text, pp 70 – 73
• The reference distribution for F0 is the Fa-1, a(n-1) distribution
• Reject the null hypothesis (equal treatment means) if
F0 > Fα ,a −1,a ( n −1)

Analysis of Variance (ANOVA): Excel Analysis:
ANOVA Single Factor
Setup data in EXCEL Spreadsheet in columns as:
15% Cotton 20% Cotton 25% Cotton 30% Cotton 35% Cotton
Tensile Strength Tensile Strength Tensile Tensile Tensile
lb/in2 lb/in2 Strength lb/in2 Strength lb/in2 Strength lb/in2
7 12 14 19 7
7 17 18 25 10
15 12 18 22 11
11 18 19 19 15
9 18 19 23 11

Analysis of Variance (ANOVA): Excel Analysis:
ANOVA Single Factor
EXCEL Data Analysis: ANOVA Single Factor

Analysis of Variance (ANOVA):

Output
EXCEL Anova: Single Factor
SUMMARY
Groups Count Sum Average Variance
15% Cotton
Tensile Strength
lb/in2 5 49 9.8 11.2
20% Cotton
Tensile Strength
lb/in2 5 77 15.4 9.8
25% Cotton
Tensile Strength
lb/in2 5 88 17.6 4.3
30% Cotton
Tensile Strength
lb/in2 5 108 21.6 6.8
35% Cotton
Tensile Strength
lb/in2 5 54 10.8 8.2

EXCEL Data Analysis: ANOVA Single Factor Output
H 0 : µ1 = µ 2 = µ 3 = ... = µ i
If H 0 true Between Sample Variation
E ( MSTr ) = E ( MSE ) = σ 2 TEST STAT =
If H 0 false
E ( MSTr ) > E ( MSE ) = σ 2 MSTr
f =
ANOVA MSE
Source of
Variation SS df MS F P-value F crit
Between
Groups 475.76 4 118.94 14.75682 9.128E-06 2.866081
Within
Groups 161.2 20 8.06
Total 636.96 24
P-Value: Probability of wrongly

rejecting the Null Fα , I −1, I ( J −1)
The Reference Distribution:

EXCEL Data Analysis: ANOVA
Manual calculations
SUMj (SUM)SQj SUM SQj
15% 20% 25% 30% 35%
Cotton Cotton Cotton Cotton Cotton
Tensile Tensile Tensile Tensile Tensile
Strength Strength Strength Strength Strength (Σxj) (Σxj)2 Σ(xj2)
7 12 14 19 7 59 3481 799
7 17 18 25 10 77 5929 1387
15 12 18 22 11 78 6084 1298
11 18 19 19 15 82 6724 1392
9 18 19 23 11 80 6400 1416
49 77 88 108 54 Σ(Σxi) 376 Σ(Σxj) Σ((Σxj)2) Σ(Σ(xj2))

(Σxi)
Σ(Σxij) Σ(Σ(xij2))
(Σxi)2 2401 5929 7744 11664 2916 Σ((Σxi2) 30654 376 28618 6292
2
Σ(xi2) 525 1225 1566 2360 616 Σ(Σ(xi )) 6292

EXCEL Data Analysis: ANOVA
Manual calculations
Manual calculations
SS - sum of MS = SS/df F statistic =
Source of SS - sum of squares estimate of MSBG/MSwithin
Variation squares SS Calculation df sigma G
Between as = [(5 x30654) -

Groups =(r( ΣΣxi )-(ΣΣxi) )/n
2 2
475.76 (376^2)]/25 4 118.94
Within as =
Groups Error 161.2 20 8.06 118.94/8.06 =
as = [(25 x 6292) -
Total =(n( ΣΣxij2)-(ΣΣxij)2)/n 636.96 (376^2)]/25 24 14.76

STAT EASE Design Expert: ANOVA Single Factor
General Factorial Design

STEP 1
Press
continue ...

STEP 2
Press
continue ...

STEP 3
Define #
replicates
Press
continue ...

STEP 4
# and
Name
response
Press
continue ...

STEP 5
Run experiment
in the defined
random order
and
measure/input
responses
Press
continue ...

STEP 6
Design Expert
Analysis

STEP 7
Design Expert
Analysis
ANOVA

STEP 8
Design Expert
Analysis

STEP 9
Design Expert
Analysis
Effects
M for Model
e for error

STEP 10 Design Expert Analysis ANOVA
same table as before in EXCEL

STEP 10 Design Expert Analysis ANOVA Terms
Model: Terms estimating factor effects. For 2-level factorials: those that "fall off" the normal probability line
of the effects plot.
Sum of Squares: Total of the sum of squares for the terms in the model, as reported in the Effects List for
factorials and on the Model screen for RSM, MIX and Crossed designs.
DF: Degrees of freedom for the model. It is the number of model terms, including the intercept, minus one.
Mean Square: Estimate of the model variance, calculated by the model sum of squares divided by model degrees
of freedom.
F Value: Test for comparing model variance with residual (error) variance. If the variances are close to the
same, the ratio will be close to one and it is less likely that any of the factors have a significant effect on the
response. Calculated by Model Mean Square divided by Residual Mean Square.
Probe > F: Probability of seeing the observed F value if the null hypothesis is true (there is no factor effect).
Small probability values call for rejection of the null hypothesis. The probability equals the proportion of the
area under the curve of the F-distribution that lies beyond the observed F value. The F distribution itself is
determined by the degrees of freedom associated with the variances being compared.
(In "plain English", if the Probe>F value is very small (less than 0.05) then the terms in the
model have a significant effect on the response.)

STEP 10 Design Expert Analysis Information in Help System
ANOVA Terms
Pure Error: Amount of variation in the response in replicated design points.
Sum of Squares: Pure error sum of squares from replicated points.
DF: The amount of information available from replicated points.
Mean Square: Estimate of pure error variance.
Cor Total: Totals of all information corrected for the mean.

Sum of Squares: Sum of the squared deviations of each point from the mean.
DF: Total degrees of freedom for the experiment, minus one for the mean.

STEP 10 Design Expert Analysis ANOVA

STEP 10 Design Expert Analysis ANOVA Information in Help System
ANOVA Terms
Next you see a collection of summary statistics for the model:
Std Dev: (Root MSE) Square root of the residual mean square. Consider this to be an estimate
of the standard deviation associated with the experiment.
Mean: Overall average of all the response data.

C.V.: Coefficient of Variation, the standard deviation expressed as a percentage of the mean.
Calculated by dividing the Std Dev by the Mean and multiplying by 100.
PRESS: Predicted Residual Error Sum of Squares – A measure of how the model fits each point
in the design. The PRESS is computed by first predicting where each point should be from a
model that contains all other points except the one in question. The squared residuals (difference
between actual and predicted values) are then summed.
R-Squared: A measure of the amount of variation around the mean explained by the model.
1-(SSresidual / (SSmodel + SSresidual))

ANOVA Terms
Summary statistics for the model continued:
Adj R-Squared: A measure of the amount of variation around the mean explained by the model,
adjusted for the number of terms in the model. The adjusted R-squared decreases as the number of
terms in the model increases if those additional terms don’t add value to the model.
1-((SSresidual / DFresidual) / ((SSmodel + SSresidual) / (DFmodel + DFresidual)))

Pred R-Squared: A measure of the amount of variation in new data explained by the model.
1-(PRESS / (SStotal-SSblock)
The predicted r-squared and the adjusted r-squared should be within 0.20 of each other. Otherwise
there may be a problem with either the data or the model. Look for outliers, consider
transformations, or consider a different order polynomial.

ANOVA Terms
Summary statistics for the model continued:

STEP 10 Design Expert Analysis ANOVA Scroll down





Model Adequacy Checking in the ANOVA
Text reference, Section 3-4, pg. 76
• Checking assumptions is important
• Normality
• Constant variance
• Independence
• Have we fit the right model?

Model Adequacy Checking in the ANOVA
STEP 10 Design Expert Analysis ANOVA Diagnostics
Residuals
• Examination of residuals
(see text, Sec. 3-4, pg. 76)
eij = yij − yˆij
= yij − yi.
• Design-Expert generates
the residuals
• Residual plots are very
useful
• Normal probability plot
of residuals

STEP 10 Design Expert Analysis ANOVA Model Graphs

Other Important Residual Plots
5.2 5.2
2.95 2.95
2
R es iduals
R es iduals
0.7 0.7
2
2
-1.55 -1.55
2
2
2
-3.8
-3.8
1 4 7 10 13 16 19 22 25
9.80 12.75 15.70 18.65 21.60
Predicted R un N um ber

Post-ANOVA Comparison of Means
• The analysis of variance tests the hypothesis of equal
treatment means
• Assume that residual analysis is satisfactory
• If that hypothesis is rejected, we don’t know which specific
means are different
• Determining which specific means differ following an
ANOVA is called the multiple comparisons problem
• There are lots of ways to do this…see text, Section 3-5, pg. 86
• We will use pairwise t-tests on means…sometimes called
Fisher’s Least Significant Difference (or Fisher’s LSD)
Method

Graphical Comparison of Means
Text, pg. 89

For the Case of Quantitative Factors, a
Regression Model is often Useful
Response:Strength
ANOVA for Response Surface Cubic Model
Analysis of variance table [Partial sum of squares]
Sum of Mean F
Source Squares DF Square Value Prob > F
Model 441.81 3 147.27 15.85 < 0.0001
A 90.84 1 90.84 9.78 0.0051
A2 343.21 1 343.21 36.93 < 0.0001
A3 64.98 1 64.98 6.99 0.0152
Residual 195.15 21 9.29
Lack of Fit 33.95 1 33.95 4.21 0.0535
Pure Error 161.20 20 8.06
Cor Total 636.96 24
Coefficient Standard 95% CI 95% CI

Factor Estimate DF Error Low High VIF
Intercept 19.47 1 0.95 17.49 21.44
A-Cotton % 8.10 1 2.59 2.71 13.49 9.03
A2 -8.86 1 1.46 -11.89 -5.83 1.00
A3 -7.60 1 2.87 -13.58 -1.62 9.03

The Regression Model
Final Equation in Terms of 25
Actual Factors:
%
20.5
Strength = +62.61143
2 2
-9.01143* Cotton Weight 2 2
Strength
% +0.48143 * Cotton 16
Weight %^2 -7.60000E-

003 * Cotton Weight %^3
2
11.5
2
This is an empirical model of

the experimental results
7 2
15.00 20.00 25.00 30.00 35.00
A: C otton Weight %

Anova - Example - P60 - 69

Uploaded by

Copyright:

Available Formats

You might also like

Anova - Example - P60 - 69

Uploaded by

Document Information

Original Description:

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

Anova - Example - P60 - 69

Uploaded by

Copyright:

Available Formats

What If There Are More Than

Two Factor Levels?

• Comparing more that two factor levels…the analysis of

ANOVA_EXAMPLE P60 Brainerd 1

ANOVA_EXAMPLE P60 Brainerd 3

•The engineer decides to test specimens at five levels of cotton weight

ANOVA_EXAMPLE P60 Brainerd 4

ANOVA_EXAMPLE P60 Brainerd 5

ANOVA_EXAMPLE P60 Brainerd 10

COTTON% Experimental Run #

ANOVA_EXAMPLE P60 Brainerd 11

ANOVA_EXAMPLE P60 Brainerd 15

Input PROCESS Output

ANOVA_EXAMPLE P60 Brainerd 16

ANOVA_EXAMPLE P60 Brainerd 17

ANOVA_EXAMPLE P60 Brainerd 18

ANOVA_EXAMPLE P60 Brainerd 20

ANOVA_EXAMPLE P60 Brainerd 22

Level Replicates Mean SD

Level Replicates Mean SD

Mean Square Treatment ≡ Between Sample Variation

Mean Square Error ≡ Within Sample Variation

ANOVA_EXAMPLE P60 Brainerd 26

• Does changing the

ANOVA_EXAMPLE P60 Brainerd 28

ANOVA_EXAMPLE P60 Brainerd 30

Setup data in EXCEL Spreadsheet in columns as:

ANOVA_EXAMPLE P60 Brainerd 31

EXCEL Data Analysis: ANOVA Single Factor

ANOVA_EXAMPLE P60 Brainerd 32

ANOVA_EXAMPLE P60 Brainerd 33

EXCEL Anova: Single Factor

ANOVA_EXAMPLE P60 Brainerd 34

P-Value: Probability of wrongly

ANOVA_EXAMPLE P60 Brainerd 36

49 77 88 108 54 Σ(Σxi) 376 Σ(Σxj) Σ((Σxj)2) Σ(Σ(xj2))

ANOVA_EXAMPLE P60 Brainerd 37

Between as = [(5 x30654) -

ANOVA_EXAMPLE P60 Brainerd 38

General Factorial Design

ANOVA_EXAMPLE P60 Brainerd 39

General Factorial Design

ANOVA_EXAMPLE P60 Brainerd 40

General Factorial Design

ANOVA_EXAMPLE P60 Brainerd 41

General Factorial Design

ANOVA_EXAMPLE P60 Brainerd 42

ANOVA_EXAMPLE P60 Brainerd 43

ANOVA_EXAMPLE P60 Brainerd 44

ANOVA_EXAMPLE P60 Brainerd 45

ANOVA_EXAMPLE P60 Brainerd 46

ANOVA_EXAMPLE P60 Brainerd 47

ANOVA_EXAMPLE P60 Brainerd 48

ANOVA_EXAMPLE P60 Brainerd 49

Cor Total: Totals of all information corrected for the mean.

ANOVA_EXAMPLE P60 Brainerd 50

STEP 10 Design Expert Analysis ANOVA