Analysis of Variance
ANOVA
Robust System Design
16.881
Session #7
MIT
Proposed Schedule Changes
? Switch lecture
?No quiz
– Informal (ungraded) presentation of term project ideas
? Read Phadke ch. 7 -- Construction Orthogonal
Arrays
– Quiz on ANOVA
– Noise experiment due
Robust System Design
16.881
Session #7
MIT
Learning Objectives
? Introduce hypothesis testing
? Introduce ANOVA in statistic practice
? Introduce ANOVA as practiced in RD
? Compare to ANOM
? Get some practice applying ANOVA in RD
? Discuss / compare / contrast
Robust System Design
16.881
Session #7
MIT
Hypothesis Testing
A technique that uses sample data from a
population to come to reasonable
conclusions with a certain degree of
confidence
Robust System Design
16.881
Session #7
MIT
Hypothesis Testing Terms
? Null Hypothesis (H
o
) -- The hypothesis to
be tested (accept/reject)
? Test statistic -- A function of the
parameters of the experiment on which you
base the test
? Critical region -- The set of values of the
test statistic that lead to rejection of H
o
Robust System Design
16.881
Session #7
MIT
Hypothesis Testing Terms (cont.)
? Level of significance (α) -- A measure of
confidence that can be placed in a result not
merely being a matter of chance
? p value -- The smallest level of significance at
which you would reject H
o
Robust System Design
16.881
Session #7
MIT
Robust System Design
Session #7
MIT16.881
Comparing the
Variance of Two Samples
? Null Hypothesis -- H
o
:
? Test Statistic --
? Acceptance criteria --
? Assumes independence & normal dist.
r=
2
1
σ
σ
F
1
r
2
Var X1(
Var X2(
.
2
1
5.0)2,1,(
α?
<?ddFpF
)
)
F Distribution
? Three arguments
– d1 (numerator DOF)
– d2 (denominator DOF)
– x (cutoff)
F(x,d1,d2)
Γ
d1 d2
2
d1
d1
2
.
d2
d2
2
.
Γ
d1
2
Γ
d2
2
.
x
d1
2
1
d1 x
.
d2(
d1 d2
2
.
for x > 0
x
)
Robust System Design
16.881
Session #7
MIT
Rolling Dice
? Population 1 -- Roll one die
? Population 2 -- Roll two die
? Go to excel sheet “dice_f_test.xls”
Robust System Design
16.881
Session #7
MIT
One-way ANOVA
? Null Hypothesis -- H
o
:
μ
1
=μ
2
=μ
3
= L
? Test Statistic --
F
SSB
SSW
dfB
dfW
? Acceptance criteria --
pF(F , dfB , dfW) < (1 α)
? Assumes independence & normal dist.
Robust System Design
16.881
Session #7
MIT
ANOVA & Robust Design
Noise Factors
H: This noise
Product / Process
Response
Signal Factor
factor affects the
mean
H: Factor setting A1 is
more robust than factor
Optimize robustness
setting A2
Control Factors
Robust System Design
16.881
Session #7
MIT
ANOVA and the Noise
Experiment
? Did the noise factors we experimented with
really have an effect on mean?
? Switch to Excel sheet
“catapult_L4_static_anova.xls”
Robust System Design
16.881
Session #7
MIT
Why Test This Hypothesis?
? Factor setting PP3 is more robust than setting PP1
? Phadke -- “In Robust Design, we are not
concerned with such probability statements, we
use the F ratio for only qualitative understanding
of the relative factor effects”
Factor Effects on the S/N Ratio
10
11
12
13
14
15
16
17
SP1 SP3 DA1 DA3
CU
P
1
CU
P3
PP
1
P
P3
S/
N Ra
t
i
o
(
d
B)
Robust System Design
16.881
Session #7
MIT
Analysis of Variance (ANOVA)
? ANOVA helps to resolve the relative
magnitude of the factor effects compared to
the error variance
? Are the factor effects real or just noise?
? I will cover it in Lecture 7.
? You may want to try the Mathcad “resource
center” under the help menu
Robust System Design
16.881
Session #7
MIT
Additive Model
? Assume each parameter affects the response
independently of the others
η( A
i
, B
j
, C
k
, D
i
) =μ+ a
i
+ b
j
+ c
k
+ d
i
+ e
A: Stop Pin B: Draw Angle C: Cup Position D: Post Pin Mean Distance Std Deviation Variance S/N Ratio
A1 B1 C1 D1 16.9 3.0 8.8 15.1
A1 B2 C2 D2 46.6 8.1 65.7 15.2
A1 B3 C3 D3 91.9 13.4 178.5 16.7
A2 B1 C2 D3 25.8 5.8 34.1 12.9
A2 B2 C3 D1 49.2 11.9 141.6 12.3
A2 B3 C1 D2 67.2 8.7 75.2 17.8
A3 B1 C3 D2 18.1 6.4 41.5 9.0
A3 B2 C1 D3 45.9 8.7 76.3 14.4
A3 B3 C2 D1 53.0 11.2 125.1 13.5
GRAND MEANS 46.1 8.6 83.0 14.1
Robust System Design
16.881
Session #7
MIT
Analysis of Means (ANOM)
? Analyze the data to discover m
A1
, a
i
...
Factor Effects on the S/N Ratio
10
11
12
13
14
15
16
17
SP
1
SP
3
D
A1
D
A
3
C
U
P
1
C
U
P
3
PP
1
PP
3
S
/
N Ra
t
i
o
(
d
B)
Robust System Design
16.881
Session #7
MIT
Analysis of Variance (ANOVA)
? Analyze data to understand the relative
contribution of control factors compared to
“error variance”
Factor Effects on the S/N Ratio
10
11
12
13
14
15
16
17
S
P
1
SP
3
DA
1
D
A3
C
UP
1
C
UP
3
PP
1
P
P3
S/
N Rat
i
o
(
d
B)
Robust System Design
16.881
Session #7
MIT
Breakdown of Sum Squares
GTSS
SS due
to mean
Total SS
SS due
to factor A
SS due
to factor B
SS due
to error
etc.
Robust System Design
16.881
Session #7
MIT
Breakdown of DOF
Robust System Design
Session #7
MIT16.881
n
1
SS due
to mean
n-1
(# levels) -1
factor A
n = Number of η values
(# levels) -1
factor B
DOF for
error
etc.
Computation of Sum of Squares
n
? Grand total sum of squares GTSS =
∑
η
i
2
i=1
? Sum of squares due to mean
= nμ
2
n
? Total sum of squares
=
∑
(η
i
?μ)
2
i=1
? Sum of squares due to a factor
= replication# [(m
A1
?μ)
2
+ (m
A2
?μ)
2
+ (m
A3
?μ)
2
]
? Sum of squares due to error
– Zero with no replicates
– Estimated by “pooling”
Robust System Design
16.881
Session #7
MIT
Pooling
? Provides an estimate of error without empty
columns or replicates
? Procedure
– Select the bottom half of the factors (in terms of
contribution to Total SS)
Robust System Design
16.881
Session #7
MIT
F-statistic
?
?
?
Error variance =
sum of squares due to error
degrees of freedom for error
F =
mean square for factor
Error variance
SS for factor
mean square for factor =
DOF for factor
– F=1 Factor effect is on par with the error
– F=2 The factor effect is marginal
– F>4 The factor effect is substantial
Robust System Design
16.881
Session #7
MIT
Confidence Intervals
for Factor Effects
? Phadke
– Variance in a
i
is error variance / replication #
– 95% confidence interval for factor effects is
two standard deviations in a
i
? How does one interpret this value?
Robust System Design
16.881
Session #7
MIT
Example
Catapult Experiment
? Switch to Excel “Catapult_L9_2.xls”
A: Stop Pin B: Draw Angle C: Cup Position D: Post Pin Mean Distance Std Deviation Variance S/N Ratio
A1 B1 C1 D1 16.9 3.0 8.8 15.1
A1 B2 C2 D2 46.6 8.1 65.7 15.2
A1 B3 C3 D3 91.9 13.4 178.5 16.7
A2 B1 C2 D3 25.8 5.8 34.1 12.9
A2 B2 C3 D1 49.2 11.9 141.6 12.3
A2 B3 C1 D2 67.2 8.7 75.2 17.8
A3 B1 C3 D2 18.1 6.4 41.5 9.0
A3 B2 C1 D3 45.9 8.7 76.3 14.4
A3 B3 C2 D1 53.0 11.2 125.1 13.5
GRAND MEANS 46.1 8.6 83.0 14.1
Robust System Design
16.881
Session #7
MIT
Homework
? Grades are exceptionally high
? Some are spending vast amounts of time
? This represents 20% of the final grade
Mean 94.2 101.0 96.5 95.4
Standard deviation 2.7 4.4 6.3 1.5
Maximum 98 109 101 97.3
Robust System Design
16.881
Session #7
MIT
Quizzes
? Some consistently score high
? Others struggling, but learning
? Remember, this is only 10%
Quiz #1 Quiz #2 Quiz #3 Quiz #4
Mean 74.3 83.4 77.5 82.8
Standard deviation 19.3 10.4 22.3 17.6
Maximum 100 100 100 110
Robust System Design
16.881
Session #7
MIT
Next Steps
? Hand in homework #5
? Homework #7 due on Lecture 10.
? Next session tomorrow
– Present your ideas for a term project
? Following session
– Quiz on ANOVA
– Homework #6 (Noise Exp.) due
– Constructing orthogonal arrays (read ch. 7)
Robust System Design
16.881
Session #7
MIT