2-11 ANOVA Analysis of Variance
2-11 ANOVA Analysis of Variance
2-11 ANOVA Analysis of Variance
Analysis Of Variance
MQPM
© 2001 ConceptFlow 1
Module Objectives
MQPM
© 2001 ConceptFlow 2
Why Learn ANOVA?
ANOVA
• Performs hypothesis testing for two or more means
• Evaluates several PIVs
• Handles multiple levels
• Shows sources of process variation
• Generates an underlying variability estimate
MQPM
© 2001 ConceptFlow 3
What Is ANOVA?
MQPM
© 2001 ConceptFlow 4
When To Use ANOVA
MQPM
© 2001 ConceptFlow 7
Process Variation: Real Or Random?
• Ho: …n
• Ha: at least one mean is different
• p-value will be used to evaluate the hypothesis
MQPM
© 2001 ConceptFlow 10
Looking For Variation
MQPM
© 2001 ConceptFlow 11
Sources Of Variation
n
wee
t
B e V Vendor B
V Vendor C
Total
Within
Variation
V Vendor A V Vendor D
MQPM
© 2001 ConceptFlow 13
The ANOVA Table Sources Of Variation
Between
Within
Sources
of
Variation
MQPM
© 2001 ConceptFlow 14
The ANOVA Table Degrees Of Freedom
Between
Within
Sources
of
Variation
Degrees of
Freedom per
Source
MQPM
© 2001 ConceptFlow 15
The ANOVA Table Sum of Squares
Between
Within
Sources Sum of
of Squares of
Variation Residuals per
Source
Degrees of
Freedom per
Source
MQPM
© 2001 ConceptFlow 16
Summarizing The Variance
yij y..
k n k k n
2
ni yi. y..
2
yij yi. 2
i 1 j 1 i 1 i 1 j 1
Source SS %
SSBetween 49.70 27.2%
SSWithin 132.99 72.8%
SSTotal 182.69
Between
Within
Sources Sum of
of Squares of
Variation Residuals per
Source
Degrees of Estimate of
Freedom per Source
Source Variance
SS/df
MQPM
© 2001 ConceptFlow 19
Mean Square Estimate Of Within Variance
SS Within
MS Within where df = N – k
df N = total number of samples
k = number of factors
132.99
MS Within 4.43 df = 34 – 4 = 30
30
MQPM
© 2001 ConceptFlow 20
Mean Square Estimate Of Between Variance
SSBetween
MSBetween where df = k – 1
df
Four vendors: df = 4 – 1 = 3
49.70
MSBetween 16.56
3
MQPM
© 2001 ConceptFlow 21
The ANOVA Table F-Ratio
Between
Within
Degrees of Estimate of
Freedom per Source
Source Variance
SS/df
MQPM
© 2001 ConceptFlow 22
Significance Of Sources
MQPM
© 2001 ConceptFlow 23
The ANOVA Table P-Value
Between
Within
MQPM
© 2001 ConceptFlow 24
ANOVA Table Exercise
MQPM
© 2001 ConceptFlow 25
ANOVA In Minitab TM
MQPM
© 2001 ConceptFlow 26
Setting Up The Data In Minitab TM
MQPM
© 2001 ConceptFlow 28
Testing Data For Normality
p=0.919 p=0.188
p=0.365 p=0.910
Tool Bar Menu > Stat > ANOVA > Test for Equal Variances
Bartlett’s Test is for normal data
Levine’s Test is for non-normal data
Levene's Test
VendorC
Test Statistic: 2.405
P-Value : 0.087
VendorD
0 1 2 3 4 5 6 7 8
Tool Bar Menu > Stat > ANOVA > Main Effects Plot
MQPM
© 2001 ConceptFlow 31
The Main Effects Plot
Tool Bar Menu > Stat > ANOVA > One Way…
MQPM
© 2001 ConceptFlow 33
The One Way ANOVA
MQPM
© 2001 ConceptFlow 35
Residuals Versus The Order Of The Data
Minitab Calculations
TM
Manual Calculations
94.56
93.76 96.44 95.20 93.20
-0.80 1.88 0.64 -1.36
Vendor A A Vendor B C Vendor C A Vendor D
91.4 -2.4 99.3 2.9 92.8 -2.4 94.4
94.6 0.8 93.7 -2.7 96.4 1.2 92.8
92.6 -1.2 99.1 2.7 96.0 0.8 90.8
95.0 1.2 99.0 2.6 94.0 -1.2 93.2
92.2 -1.6 92.8 -3.6 92.8 -2.4 95.2
97.0 3.2 96.7 0.3 95.6 0.4 93.2
89.4 -4.4 94.5 -1.9 96.8 1.6 92.0
95.4 1.6 97.2 2.0 94.0
93.4 -0.4 95.2 0.0
96.6 2.8
MQPM
© 2001 ConceptFlow 39
Normal Probability Plot Of The Residuals
MQPM
© 2001 ConceptFlow 44
Sample Size For ANOVA
Tool Bar Menu > Stat > Power and Sample Size > ANOVA
• Determine the sample needed to detect a mean shift of 3.2 on a
process with a variance of 4 when conducting a single variable test
at 3 different settings. Use an alpha of 5% and beta of 20%
MQPM
© 2001 ConceptFlow 45
Sample Size For ANOVA
MQPM
© 2001 ConceptFlow 46
Sample Size For ANOVA
Continued
MQPM
© 2001 ConceptFlow 48
Key Learning Points
•
•
•
•
•
•
MQPM
© 2001 ConceptFlow 49
Objectives Review
MQPM
© 2001 ConceptFlow 50
Appendix
MQPM
© 2001 ConceptFlow 51
Decomposing The Data
MQPM
© 2001 ConceptFlow 52
Summation Notation
Given Find
Reading Group 1 n
1 9.8 1) yj ?
2 12.0 j1
3 17.3 n
4
5
10.0
10.4
yj
2) j1 What is
6 12.5 ? this value
72.00 n called?
Group the data by source (factor) and calculate the factor means
93.76 96.44 95.20 ? Factor Mean
Day Vendor A Vendor B Vendor C Vendor D n
1 91.4 99.3 92.8 94.4 y ij
j 1
2
3
94.6
92.6
93.7
99.1
96.4
96.0
92.8
90.8 i. y
4 95.0 99.0 94.0 93.2
n
5 92.2 92.8 92.8 95.2 y1. 93.76 n1 10
6 97.0 96.7 95.6 93.2
7 89.4 94.5 96.8 92.0 y 2. 96.44 n2 7
8 95.4 97.2 94.0 y 3. 95.20 n3 9
9 93.4 95.2
10 96.6 y 4. ? n4 ?
MQPM
© 2001 ConceptFlow 55
Double Summation Notation
2 4
y ij y11 y12 y13 y14
i 1 j 1
y 21 y 22 y 23 y 24
Given Find
Reading Group 1 Group 2 k n
1
2
9.8
12.0
12.7
13.6 y ij
3 17.3 16.4 i 1 j 1
4 10.0 14.3
5 10.4 13.0
6 12.5 12.0
Group the data by source (factor) and calculate the overall (grand)
mean
94.56 Grand Mean y..
93.76 96.44 95.20 93.20 Factor Mean y
i.
Day Vendor A Vendor B Vendor C Vendor D
1 91.4 99.3 92.8 94.4 k ni
2 94.6 93.7 96.4 92.8
j 1 i 1
yij
3 92.6 99.1 96.0 90.8
y.. k
4 95.0 99.0 94.0 93.2
5 92.2 92.8 92.8 95.2
j 1
nj
6 97.0 96.7 95.6 93.2
7 89.4 94.5 96.8 92.0
8 95.4 97.2 94.0 For this data set,
9 93.4 95.2 what are the values
10 96.6 of k and n?
MQPM
© 2001 ConceptFlow 58
Determining The Factor Contributions
B C
y A D
MQPM
© 2001 ConceptFlow 63
Understanding Within Residuals
ij y ij y i
92.2 -1.6 92.8 -3.6 92.8 -2.4 95.2
97.0 3.2 96.7 0.3 95.6 0.4 93.2
89.4 -4.4 94.5 -1.9 96.8 1.6 92.0
95.4 1.6 97.2 2.0 94.0
93.4 -0.4 95.2 0.0
96.6 2.8
MQPM
© 2001 ConceptFlow 66
Summary Of Data Decomposing
MQPM
© 2001 ConceptFlow 67
Analyzing The ANOVA Data
MQPM
© 2001 ConceptFlow 68
Working With The ANOVA Data
MQPM
© 2001 ConceptFlow 69
Total Process Variation
Total process variance is the sum of the squares of the overall residuals
k n
SS Total ( y ij y..) 2
i 1 j 1
Within process variance is the sum of the squares of the within residuals
k n
SS Within ( yij yi.)2
i 1 j 1
2 2 2 2
A A B B C C D D
-2.4 5.570 -2.9 8.165 2.4 5.760 -1.2 1.440
0.8 0.706 2.8 7.685 -1.2 1.440 0.4 0.160
-1.2 1.346 -2.6 6.866 -0.8 0.640 2.4 5.760 k n
1.2 1.538 -2.5 6.397 1.2 1.440 0.0 0.000 SS Within ( y ij y i .) 2
-1.6 2.434 3.6 12.920 2.4 5.760 -2.0 4.000 i 1 j 1
3.2 10.498 -0.3 0.080 -0.4 0.160 0.0 0.000
-4.4 19.010 1.9 3.699 -1.6 2.560 1.2 1.440
1.6 2.690 SSB 45.812 -2 4.000 -0.8 0.640
-0.4 0.130 0 0.000 SSD 13.440
2.8 8.066 SSC 21.760
SSA 51.984
MQPM
© 2001 ConceptFlow 73
Between Factor Variation
2 2
s
y n
MQPM
© 2001 ConceptFlow 74
Deriving Factor Variation
The estimate of the variance, , is related to the sample variance for
each source
2 2
2 2
s Solving for 2
s n
y n y
2
s
kj1 y . y..
j
2
y k 1
MQPM
© 2001 ConceptFlow 75
Between Process Variation
The between process variance is the sum of the squares of the sample-
size-weighted contribution factors squared
k k
2 2
SSBetween s ni ni yi. y..
i 1 yi i 1
Factor Contribution
Residual from Grand Sy2 n Sy2*n
Mean k 2
-0.80 0.643 10 6.43 SSBetween ni yi. y..
i 1
1.88 3.539 7 24.77
0.64 0.407 9 3.67
-1.36 1.854 8 14.84
Calc/ProbabilityDistributions/F
MQPM
© 2001 ConceptFlow 78
Class Exercise
MQPM
© 2001 ConceptFlow 79
On-Time Delivery Problem
MQPM
© 2001 ConceptFlow 81