Sie sind auf Seite 1von 13

IE563 Homework #2 Solutions

Problem 1
(a)
Response = tensile strength
Factor = mixing technique
4 factor levels
5 replicates
(b)
Mixing Technique
1
2
3
4

source of variation
mixing technique
error
total

1
352
639
531
617

Tensile Strength (lb/in2) measurements


2
3
4
508
426
328
558
373
427
418
440
361
584
629
595

SS
75303.35
111689.6
186992.95

df
MS
3 25101.117
16
6980.6
19

F
3.5958

5
525
406
421
481

P
0.03697

Yi.
2139
2403
2171
2906
Y..
Y..2

9619
92525161

Yij2

4813251

SSTotal
SSFactor
SSError

186992.95
75303.35
111689.60

Yi.2
4575321
5774409
4713241
8444836

Since F = 3.5958 > F 0.05 = 3.24, we reject H0: 1= 2= 3= 4 i.e. we do not believe the
mean response is the same for all factor levels. We believe at least one factor level has a
mean response significantly different than the mean responses at the other factor levels.

Do mixing techniques affect the tensile strength of the cement? YES

Problem 2
(a)
General Linear Model: strength versus mix technique
Method
Factor coding

(-1, 0, +1)

Factor Information
Factor
Type
mix technique Fixed

Levels
4

Analysis of Variance
Source
DF Adj SS
mix technique
3
75303
Error
16 111690
Total
19 186993
Model Summary
S
R-sq
83.5500 40.27%

R-sq(adj)
29.07%

Values
1, 2, 3, 4
Adj MS
25101
6981

F-Value
3.60

P-Value
0.037

R-sq(pred)
6.67%

Since p = 0.037 < = 0.05, we reject H0: 1= 2= 3= 4 i.e. we do not believe the mean
response is the same for all factor levels. We believe at least one factor level has a mean
response significantly different than the mean responses at the other factor levels.
Do mixing techniques affect the tensile strength of the cement? YES

(b)
Tukey Pairwise Comparisons: Response = strength, Term = mix technique
Grouping Information Using the Tukey Method and 95% Confidence
mix
technique N
Mean Grouping
4
5 581.2 A
2
5 480.6 A
B
3
5 434.2 A
B
1
5 427.8
B
Means that do not share a letter are significantly different.
Tukey Simultaneous Tests for Differences of Means
Difference
of mix
technique
Levels
2 - 1
3 - 1
4 - 1
3 - 2
4 - 2
4 - 3

Difference
of Means
52.8
6.4
153.4
-46.4
100.6
147.0

SE of
Difference
52.8
52.8
52.8
52.8
52.8
52.8

Simultaneous
95% CI
( -98.5, 204.1)
(-144.9, 157.7)
(
2.1, 304.7)
(-197.7, 104.9)
( -50.7, 251.9)
( -4.3, 298.3)

T-Value
1.00
0.12
2.90
-0.88
1.90
2.78

Adjusted
P-Value
0.752
0.999
0.046
0.816
0.265
0.058

Choose mixing techniques 4, 3, or 2 to maximize tensile strength.


Choose mixing technique 1, 2, or 3 to minimize tensile strength.

Problem 3
(a)
Response: density
Factor: oven temperature
# Factor Levels: 6 (1800, 1900, 1950, 2000, 2050, 2100)
# replicates at temperature of 1800: 6
# replicates at temperature of 1900: 7
# replicates at temperature of 1950: 4
# replicates at temperature of 2000: 8
# replicates at temperature of 2050: 5
# replicates at temperature of 2100: 10

(b)
General Linear Model: Brick Density versus Oven Temperature
Method
Factor coding

(-1, 0, +1)

Factor Information
Factor
Type
Oven Temperature Fixed
Analysis of Variance
Source
DF
Oven Temperature
5
Error
34
Total
39

Levels
6
Adj SS
6928
9479
16407

Values
1800, 1900, 1950, 2000, 2050, 2100
Adj MS
1385.7
278.8

F-Value
4.97

P-Value
0.002

Since p = 0.002 < = 0.05, we reject H0 i.e. we do not believe the mean response is the
same for all factor levels.
Does oven temperature affect the density of bricks? YES

(c)
Tukey Pairwise Comparisons: Response = Brick Density, Term = Oven
Temperature
Grouping Information Using the Tukey Method and 95% Confidence
Oven
Temperature
2000
1950
1900
2100
2050
1800

N
8
4
7
10
5
6

Mean
143.325
133.500
125.657
118.010
111.640
103.683

Grouping
A
A
B
A
B
B
B
B

Means that do not share a letter are significantly different.


Tukey Simultaneous Tests for Differences of Means
Difference
of Oven
Temperature
Levels
1900 - 1800
1950 - 1800
2000 - 1800
2050 - 1800
2100 - 1800
1950 - 1900
2000 - 1900
2050 - 1900
2100 - 1900
2000 - 1950
2050 - 1950
2100 - 1950
2050 - 2000
2100 - 2000
2100 - 2050

Difference
of Means
21.97
29.8
39.64
8.0
14.33
7.8
17.67
-14.02
-7.65
9.8
-21.9
-15.49
-31.69
-25.32
6.37

SE of
Difference
9.29
10.8
9.02
10.1
8.62
10.5
8.64
9.78
8.23
10.2
11.2
9.88
9.52
7.92
9.15

Simultaneous
95% CI
( -6.07, 50.02)
( -2.7, 62.4)
( 12.41, 66.87)
( -22.6, 38.5)
(-11.71, 40.36)
( -23.8, 39.4)
( -8.42, 43.76)
(-43.54, 15.50)
(-32.49, 17.20)
( -21.0, 40.7)
( -55.7, 12.0)
(-45.32, 14.34)
(-60.43, -2.94)
(-49.23, -1.40)
(-21.24, 33.98)

T-Value
2.37
2.77
4.40
0.79
1.66
0.75
2.04
-1.43
-0.93
0.96
-1.95
-1.57
-3.33
-3.20
0.70

Adjusted
P-Value
0.197
0.088
0.001
0.968
0.565
0.974
0.340
0.707
0.936
0.927
0.390
0.624
0.024
0.033
0.981

Oven temperatures of 2000, 1950, and 1900 maximize the mean brick density at =0.05.
Temperatures of 1950, 1900, 2100, 2050, and 1800 will minimize the meant brick density
at =0.05.

Problem 4
(a)
Response = % radon released
Factor = orifice diameter
6 factor levels
5 replicates at each factor level

(b)
General Linear Model: Radon Released versus Orifice Diameter
Method
Factor coding

(-1, 0, +1)

Factor Information
Factor
Type
Orifice Diameter Fixed
Analysis of Variance
Source
DF
Orifice Diameter
5
Error
24

Levels
6
Adj SS
1963.5
477.5

Values
37, 51, 71, 102, 140, 199
Adj MS
392.70
19.90

F-Value
19.74

P-Value
0.000

Total
Model Summary
S
R-sq
4.46057 80.44%

29

2441.0

R-sq(adj)
76.36%

R-sq(pred)
69.43%

Since p=0.000 < = 0.05, we reject H0, i.e. we do not believe the mean response is the
same for all factor levels. We believe at least one factor level has a mean response
significantly different than the mean responses at the other factor levels.
Does the size of the orifice affect the mean percentage of radon released? YES

(c)

All ANOVA assumptions appear to be met:


Residuals are Normally distributed with mean 0 (see normal probability plot).
Residuals are independently distributed (see residuals vs observation order plot).
Residuals have constant variance (see residuals vs fitted values plot).

(d)
Tukey Pairwise Comparisons: Response = Radon Released, Term = Orifice
Diameter
Grouping Information Using the Tukey Method and 95% Confidence
Orifice
Diameter N
Mean Grouping
51
5 77.62 A
37
5 76.56 A
71
5 71.74 A
B
102
5 65.08
B
140
5 63.12
B C
199
5 54.56
C
Means that do not share a letter are significantly different.
Tukey Simultaneous Tests for Differences of Means
Difference
of Orifice
Diameter
Difference
SE of
Simultaneous
Levels
of Means Difference
95% CI
51 - 37
1.06
2.82 ( -7.66,
9.78)
71 - 37
-4.82
2.82 (-13.54,
3.90)
102 - 37
-11.48
2.82 (-20.20, -2.76)
140 - 37
-13.44
2.82 (-22.16, -4.72)
199 - 37
-22.00
2.82 (-30.72, -13.28)
71 - 51
-5.88
2.82 (-14.60,
2.84)
102 - 51
-12.54
2.82 (-21.26, -3.82)
140 - 51
-14.50
2.82 (-23.22, -5.78)
199 - 51
-23.06
2.82 (-31.78, -14.34)
102 - 71
-6.66
2.82 (-15.38,
2.06)
140 - 71
-8.62
2.82 (-17.34,
0.10)
199 - 71
-17.18
2.82 (-25.90, -8.46)
140 - 102
-1.96
2.82 (-10.68,
6.76)
199 - 102
-10.52
2.82 (-19.24, -1.80)
199 - 140
-8.56
2.82 (-17.28,
0.16)
Individual confidence level = 99.50%

T-Value
0.38
-1.71
-4.07
-4.76
-7.80
-2.08
-4.45
-5.14
-8.17
-2.36
-3.06
-6.09
-0.69
-3.73
-3.03

Adjusted
P-Value
0.999
0.540
0.005
0.001
0.000
0.328
0.002
0.000
0.000
0.210
0.054
0.000
0.981
0.012
0.057

Orifice diameter of 140 or 199 mm produced the statistically significant minimum mean
radon released.

(e)
Regression Analysis: Radon Released versus Orifice Diameter
Analysis of Variance
Source
DF
Regression
1
Orifice Diameter
1
Error
28
Lack-of-Fit
4
Pure Error
24
Total
29

Adj SS
1887.17
1887.17
553.86
76.34
477.52
2441.03

Model Summary
S
R-sq
4.44756 77.31%

R-sq(adj)
76.50%

Coefficients
Term
Constant
Orifice Diameter

Coef
82.35
-0.1424

Adj MS
1887.17
1887.17
19.78
19.09
19.90

F-Value
95.40
95.40

P-Value
0.000
0.000

0.96

0.448

R-sq(pred)
74.43%
SE Coef
1.67
0.0146

T-Value
49.35
-9.77

P-Value
0.000
0.000

VIF
1.00

Regression Equation
Radon Released = 82.35 - 0.1424 Orifice Diameter

All regression assumptions appear to be met:


Residuals are Normally distributed with mean 0 (see normal probability plot).
Residuals are independently distributed (see residuals vs observation order plot).
Residuals have constant variance (see residuals vs fitted values plot).

(f)
R2 is 77.31% indicating this regression equation is a good fit to the data. I would expect
predictions made with this regression equation to be reasonably close to the actual
observations.
(g) 64.
Radon Released = 82.35 - 0.1424*(125) = 64.55

(h)
Prediction for Radon Released
Regression Equation
Radon Released = 82.35 - 0.1424 Orifice Diameter
Variable
Orifice Diameter
Fit
64.5536

SE Fit
0.890046

Setting
125
95% CI
(62.7304, 66.3768)

95% PI
(55.2625, 73.8447)

Problem 5
(a)
General Linear Model: Yield versus Temperature
Method
Factor coding

(-1, 0, +1)

Factor Information
Factor
Type
Temperature Fixed

Levels
6

Analysis of Variance
Source
DF Adj SS
Temperature
5
229.3
Error
12
106.5
Total
17
335.8

Values
250, 300, 375, 450, 500, 550
Adj MS
45.868
8.876

F-Value
5.17

P-Value
0.009

Since p=0.009 < = 0.05, we reject H0, i.e. we do not believe the mean response is the
same for all factor levels. We believe at least one factor level has a mean response
significantly different than the mean responses at the other factor levels.
Does temperature affect the mean yield? YES

All ANOVA assumptions appear to be met:


Residuals are Normally distributed with mean 0 (see normal probability plot).
Residuals are independently distributed (see residuals vs observation order plot).
Residuals have somewhat constant variance (see residuals vs fitted values plot).

(b)
Tukey Pairwise Comparisons: Response = Yield, Term = Temp
Grouping Information Using the Tukey Method and 95% Confidence
Temp
550
500
300
250
450
375

N
3
3
3
3
3
3

Mean
33.7475
28.7438
27.8104
26.7608
24.9559
22.1285

Grouping
A
A
B
A
B
A
B
B
B

Means that do not share a letter are significantly different.


Tukey Simultaneous Tests for Differences of Means
Difference of
Temp Levels
300 - 250
375 - 250
450 - 250
500 - 250
550 - 250
375 - 300
450 - 300
500 - 300

Difference
of Means
1.05
-4.63
-1.80
1.98
6.99
-5.68
-2.85
0.93

SE of
Difference
2.42
2.42
2.42
2.42
2.42
2.42
2.42
2.42

Simultaneous
95% CI
( -7.09, 9.19)
(-12.77, 3.51)
( -9.95, 6.34)
( -6.16, 10.12)
( -1.15, 15.13)
(-13.82, 2.46)
(-11.00, 5.29)
( -7.21, 9.07)

T-Value
0.43
-1.91
-0.74
0.82
2.88
-2.34
-1.18
0.39

Adjusted
P-Value
0.998
0.440
0.972
0.958
0.110
0.249
0.839
0.999

550
450
500
550
500
550
550

300
375
375
375
450
450
500

5.94
2.83
6.62
11.62
3.79
8.79
5.00

2.42
2.42
2.42
2.42
2.42
2.42
2.42

(
(
(
(
(
(
(

-2.20,
-5.31,
-1.53,
3.48,
-4.35,
0.65,
-3.14,

14.08)
10.97)
14.76)
19.76)
11.93)
16.93)
13.14)

2.45
1.17
2.73
4.79
1.56
3.63
2.06

0.214
0.844
0.140
0.005
0.635
0.032
0.365

Temperatures of 550, 300, 500, and 250 degrees produced the statistically significant
maximum mean yields.
(c)
Regression Analysis: Yield versus Temp
Analysis of Variance
Source
DF Adj SS
Regression
1
58.53
Temp
1
58.53
Error
16 276.49
Lack-of-Fit
4 170.75
Pure Error
12 105.74
Total
17 335.02
Model Summary
S
R-sq
4.15700 17.47%

R-sq(adj)
12.31%

Coefficients
Term
Coef
Constant
20.51
Temp
0.01694

SE Coef
3.85
0.00920

Adj MS
58.530
58.530
17.281
42.687
8.812

F-Value
3.39
3.39

P-Value
0.084
0.084

4.84

0.015

R-sq(pred)
0.00%
T-Value
5.33
1.84

P-Value
0.000
0.084

VIF
1.00

Regression Equation
Yield = 20.51 + 0.01694 Temp

BAD FIT FOR REGRESSION!! SEE CURVITURE IN VERSUS FITS. NEED


HIGHER ORDER TERMS. RERUN WITH TEMP^2.

10

Regression Analysis: Yield versus Temp, Temp^2


Analysis of Variance
Source
DF Adj SS
Regression
2 195.18
Temp
1 117.89
Temp^2
1 136.65
Error
15 139.84
Lack-of-Fit
3
34.10
Pure Error
12 105.74
Total
17 335.02
Model Summary
S
R-sq
3.05330 58.26%

R-sq(adj)
52.69%

Coefficients
Term
Coef
Constant
67.7
Temp
-0.2381
Temp^2
0.000320

Adj MS
97.590
117.894
136.651
9.323
11.365
8.812

F-Value
10.47
12.65
14.66

P-Value
0.001
0.003
0.002

1.29

0.323

R-sq(pred)
43.84%

SE Coef
12.7
0.0670
0.000084

T-Value
5.35
-3.56
3.83

P-Value
0.000
0.003
0.002

VIF
98.16
98.16

Regression Equation
Yield = 67.7 - 0.2381 Temp + 0.000320 Temp^2

All assumptions appear to be met:


Residuals are Normally distributed with mean 0 (see normal probability plot).
Residuals are independently distributed (see residuals vs observation order plot).
Residuals have semi-constant variance (see residuals vs fitted values plot).
o SLIGHT FUNNELING, USING BOX-COX TRANSOFRM OK (see
below)

11

**Not Required**
If student ALSO applied Box-Cox Transformation, would look like this. The remainder
of the problem is worked out without the transformation, but no points should be taken
off if student used transformed model, so long as they undid the transformation for
results.
Regression Analysis: Yield versus Temp, Temp^2
Method
Box-Cox transformation
Rounded
Estimated
95% CI for

4
3.61141
(1.02591, 6.27691)

Analysis of Variance for Transformed Response


Source
DF Adj SS
Adj MS F-Value P-Value
Regression
2 266.06 133.031
17.65
0.000
Temp
1 127.63 127.627
16.94
0.001
Temp^2
1 154.43 154.432
20.49
0.000
Error
15 113.04
7.536
Lack-of-Fit
3
36.71
12.235
1.92
0.180
Pure Error
12
76.33
6.361
Total
17 379.10
Model Summary for Transformed Response
S
R-sq R-sq(adj) R-sq(pred)
2.74518 70.18%
66.21%
59.45%
Coefficients for Transformed Response
Term
Coef
SE Coef T-Value
Constant
48.9
11.4
4.30
Temp
-0.2478
0.0602
-4.12
Temp^2
0.000340 0.000075
4.53

P-Value
0.001
0.001
0.000

VIF
98.16
98.16

Regression Equation
(Yield^-1)/(g^(-1)) = 48.9 - 0.2478 Temp + 0.000340 Temp^2
( = 4, g = 26.9962 is the geometric mean of Yield)

12

(d)
Yield = 67.7 - 0.2381*(330) + 0.000320*(330*330)= 23.975

(e)
Prediction for Yield
Regression Equation
Yield = 67.7 - 0.2381 Temp + 0.000320 Temp^2
Variable
Temp
Temp^2
Fit
23.9764

Setting
330
108900
SE Fit
1.03798

95% CI
(21.7640, 26.1888)

95% PI
(17.1027, 30.8501)

Problem 6
Power and Sample Size
One-way ANOVA
= 0.05

Assumed standard deviation = 83.5524

Factors: 1

Number of levels: 4

Maximum
Difference
6

Sample
Size
5498

Target
Power
0.9

From problem 1 ANOVA MSE

Actual Power
0.900049

The sample size is for each level.

13

Das könnte Ihnen auch gefallen