Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

0.8
0.7
0.6
0.5
y 0.4
0.3
0.2
0.1
0
-50 0 50 100 150

b) yˆ = 0.32999 + 0.00416(85) = 0.6836

c) yˆ = 0.32999 + 0.00416(90) = 0.7044

d) βˆ1 = 0.00416

11-3 a)
Regression Analysis: Rating Points versus Meters per Att

The regression equation is

y = 14.2 + 11.0 x

Predictor Coef SE Coef T P

Constant 14.185 9.089 1.56 0.129
x 11.039 1.413 7.81 0.000

S = 5.22972 R-Sq = 67.0% R-Sq(adj) = 65.9%

Analysis of Variance
Source DF SS MS F P
Regression 1 1669.0 1669.0 61.02 0.000
Residual Error 30 820.5 27.4
Total 31 2489.5

yi = β 0 + β1 xi + ε i

(204.74) 2
S xx = 1323.648 − = 13.696
32
(204.74)(2714.1)
S xy = 17516.34 − = 151.1889
32
S xy 151.1889
βˆ1 = = = 11.039
S xx 13.696
2714.1 ⎛ 204.74 ⎞
βˆ0 = − (11.039) ⎜ ⎟ = 14.187
32 ⎝ 32 ⎠

11-2
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

yˆ = 14.2 + 11x

SS E 820.5
σˆ 2 = MS E = = = 27.35
n−2 30

b) yˆ = 14.2 + 11(6.9) = 90.1

c) − βˆ1 = −11

1
d) × 10 = 0.91
11
e) yˆ = 14.2 + 11(6.59) = 86.69

There are two residuals

e = y − yˆ
e1 = 90.2 − 86.69 = 3.51
e2 = 95 − 86.69 = 8.31

11-4 a)
Regression Analysis - Linear model: Y = a+bX
Dependent variable: SalePrice Independent variable: Taxes
---------------------------------------------------------------------------
Standard T Prob.
Parameter Estimate Error Value Level
Intercept 13.3202 2.57172 5.17948 .00003
Slope 3.32437 0.390276 8.518 .00000
---------------------------------------------------------------------------

Analysis of Variance
Source Sum of Squares Df Mean Square F-Ratio Prob.
Level

11-3
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Model 636.15569 1 636.15569 72.5563

.00000
Residual 192.89056 22 8.76775
--------------------------------------------------------------------------
Total (Corr.) 829.04625 23
Correlation Coefficient = 0.875976 R-squared = 76.73 percent
Stnd. Error of Est. = 2.96104

σˆ 2 = 8.76775
If the calculations were to be done by hand, use Equations (11-7) and (11-8).
yˆ = 13.3202 + 3.32437 x

b) yˆ = 13.3202 + 3.32437(7.3) = 37.588

c) yˆ = 13.3202 + 3.32437(5.6039) = 31.9496

yˆ = 31.9496
e = y − yˆ = 28.9 − 31.9496 = −3.0496
d) All the points would lie along a 45 degree line. That is, the regression model would estimate the
values exactly. At this point, the graph of observed vs. predicted indicates that the simple linear
regression model provides a reasonable fit to the data.

Plot of Observed values versus predicted

50

45
Predicted

40

35

30

25

25 30 35 40 45 50

Observed

11-5 a)
Regression Analysis - Linear model: Y = a+bX
Dependent variable: Usage Independent variable: Temperature
----------------------------------------------------------------------------
Standard
Parameter Estimate Error T Prob.

11-4
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Intercept 129.974 0.707 183.80 0.000

Slope 7.59262 0.05798 130.95 0.000
----------------------------------------------------------------------------
Analysis of Variance
Source DF Sum of Squares Mean Square F Prob. Level
Model 1 57701 57701 7148.85 0.000
Residual 10 34 3
Total 11 57734
----------------------------------------------------------------------------
Stnd. Error of Est. = 1.83431 R-Sq = 99.9%
Correlation Coefficient = 0.9999

σˆ 2 = 3
If the calculations were to be done by hand, use Equations (11-7) and (11-8).
yˆ = 130 + 7.59 x

b) yˆ = −130 + 7.59(13) = 228.67

c) If monthly temperature increases by 0.5°C, ŷ increases by 7.59

d) yˆ = −130 + 7.59(8) = 190.72

yˆ = 190.72

e = y − yˆ = 192.70 − 190.72 = 1.98

11-6 a)
The regression equation is
MPG = 39.2 - 0.0402 Engine Displacement

Predictor Coef SE Coef T P

Constant 39.156 2.006 19.52 0.000
Engine Displacement -0.040216 0.007671 -5.24 0.000

S = 3.74332 R-Sq = 59.1% R-Sq(adj) = 57.0%

Analysis of Variance
Source DF SS MS F P
Regression 1 385.18 385.18 27.49 0.000
Residual Error 19 266.24 14.01
Total 20 651.41

σˆ 2 = 14.01
yˆ = 39.2 − 0.0402 x

11-5
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

b) yˆ = 39.2 − 0.0402(150) = 33.17

c) yˆ = 34.2956

e = y − yˆ = 41.3 − 34.2956 = 7.0044

11-7 a)

60

50
y

40

850 950 1050

x

Predictor Coef StDev T P

Constant -16.509 9.843 -1.68 0.122
x 0.06936 0.01045 6.64 0.000
S = 2.706 R-Sq = 80.0% R-Sq(adj) = 78.2%

Analysis of Variance
Source DF SS MS F P
Regression 1 322.50 322.50 44.03 0.000
Error 11 80.57 7.32
Total 12 403.08

σˆ 2 = 7.3212
yˆ = −16.5093 + 0.0693554 x

b) yˆ = 46.9509 e = 46.9509 − 46 = 0.9509

c) yˆ = −16.5093 + 0.0693554(960) = 50.07

11-8 a)

9
8
7
6
5
y

4
3
2
1
0

60 70 80 90 100
x

11-6
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Yes, a linear regression would seem appropriate, but one or two points might be outliers.

Predictor Coef SE Coef T P

Constant -10.132 1.995 -5.08 0.000
x 0.17429 0.02383 7.31 0.000
S = 1.318 R-Sq = 74.8% R-Sq(adj) = 73.4%

Analysis of Variance
Source DF SS MS F P
Regression 1 92.934 92.934 53.50 0.000
Residual Error 18 31.266 1.737
Total 19 124.200

b) σˆ 2 = 1.737 and yˆ = −10.132 + 0.17429 x

c) yˆ = 5.5541 at x = 90

11-9 a)

250

200
y

150

100

0 10 20 30 40
x

Yes, a linear regression model appears to be plausible.

Predictor Coef StDev T P

Constant 234.07 13.75 17.03 0.000
x -3.5086 0.4911 -7.14 0.000
S = 19.96 R-Sq = 87.9% R-Sq(adj) = 86.2%

Analysis of Variance
Source DF SS MS F P
Regression 1 20329 20329 51.04 0.000
Error 7 2788 398
Total 8 23117

b) σˆ 2 = 398.25 and yˆ = 234.071 − 3.50856 x

c) yˆ = 234.071 − 3.50856(30) = 128.814

d) yˆ = 156.883 e = 15.1175

11-7
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-10 a)

40

30

y
20

10

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
x

Yes, a simple linear regression model seems appropriate for these data.

Predictor Coef StDev T P

Constant 0.470 1.936 0.24 0.811
x 20.567 2.142 9.60 0.000
S = 3.716 R-Sq = 85.2% R-Sq(adj) = 84.3%

Analysis of Variance

Source DF SS MS F P
Regression 1 1273.5 1273.5 92.22 0.000
Error 16 220.9 13.8
Total 17 1494.5

b) σˆ 2 = 13.81
yˆ = 0.470467 + 20.5673 x

c) yˆ = 0.470467 + 20.5673(1) = 21.038

d) yˆ = 13.42787 e = 2.5279 for x = 0.63

11-11 a)

11-8
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Yes, a simple linear regression (straight-line) model seems plausible for this situation.

Predictor Coef SE Coef T P

Constant 18090.2 310.8 58.20 0.000
x -254.55 20.34 -12.52 0.000

S = 678.964 R-Sq = 89.7% R-Sq(adj) = 89.1%

Analysis of Variance
Source DF SS MS F P
Regression 1 72222688 72222688 156.67 0.000
Residual Error 18 8297849 460992
Total 19 80520537

b) σˆ 2 = 460992
yˆ = 18090.2 − 254.55 x

c) yˆ = 18090.2 − 254.55(20) = 12999.2

d) If there were no error, the values would all lie along the 45° line. The plot indicates age is a
reasonable regressor variable.

11-9
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-12 a)
The regression equation is
Porosity = 55.6 - 0.0342 Temperature

Predictor Coef SE Coef T P

Constant 55.63 32.11 1.73 0.144
Temperature -0.03416 0.02569 -1.33 0.241

S = 8.79376 R-Sq = 26.1% R-Sq(adj) = 11.3%

Analysis of Variance

Source DF SS MS F P
Regression 1 136.68 136.68 1.77 0.241
Residual Error 5 386.65 77.33
Total 6 523.33

yˆ = 55.63 − 0.03416 x

σˆ 2 = 77.33

b) yˆ = 55.63 − 0.03416(1400) = 7.806

c) yˆ = 4.39 e = 7.012

11-10
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

d)

The simple linear regression model doesn’t seem appropriate because the scatter plot doesn’t show a
linear relationship between the data.

11-13 a)
The regression equation is
BOD = 0.658 + 0.178 Time

Predictor Coef SE Coef T P

Constant 0.6578 0.1657 3.97 0.003
Time 0.17806 0.01400 12.72 0.000

S = 0.287281 R-Sq = 94.7% R-Sq(adj) = 94.1%

Analysis of Variance
Source DF SS MS F P
Regression 1 13.344 13.344 161.69 0.000
Residual Error 9 0.743 0.083
Total 10 14.087

yˆ = 0.658 + 0.178 x

σˆ 2 = 0.083

b) yˆ = 0.658 + 0.178(15) = 3.328

c) 0.178(3) = 0.534

d) yˆ = 0.658 + 0.178(6) = 1.726

e = y − yˆ = 1.9 − 1.726 = 0.174

11-11
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

e)

All the points would lie along the 45 degree line y = yˆ . That is, the regression model would estimate

the values exactly. At this point, the graph of observed vs. predicted indicates that the simple linear
regression model provides a reasonable fit to the data.

11-14 a)
The regression equation is
Deflection = 32.0 - 0.277 Stress level

Predictor Coef SE Coef T P

Constant 32.049 2.885 11.11 0.000
Stress level -0.27712 0.04361 -6.35 0.000

S = 1.05743 R-Sq = 85.2% R-Sq(adj) = 83.1%

Analysis of Variance

Source DF SS MS F P
Regression 1 45.154 45.154 40.38 0.000
Residual Error 7 7.827 1.118
Total 8 52.981

σˆ 2 = 1.118

11-12
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

b) yˆ = 32.05 − 0.277(64) = 14.322

c) (−0.277)(5) = −1.385
1
d) = 3.61
0.277
e) yˆ = 32.05 − 0.277(75) = 11.275 e = y − yˆ = 12.534 − 11.275 = 1.259

11-15

It’s possible to fit this data with straight-line model, but it’s not a good fit. There is a curvature shown
on the scatter plot.

a)
The regression equation is
y = 2.02 + 0.0287 x

Predictor Coef SE Coef T P

Constant 2.01977 0.05313 38.02 0.000
x 0.028718 0.003966 7.24 0.000

S = 0.159159 R-Sq = 67.7% R-Sq(adj) = 66.4%

Analysis of Variance

11-13
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Source DF SS MS F P
Regression 1 1.3280 1.3280 52.42 0.000
Residual Error 25 0.6333 0.0253
Total 26 1.9613

yˆ = 2.02 + 0.0287 x

σˆ 2 = 0.0253

b) yˆ = 2.02 + 0.0287(11) = 2.3357

c)

If the relationship between length and age was deterministic, the points would fall on the 45 degree
line y = yˆ . Because the points in this plot vary substantially from this line, it appears that age is not a

reasonable choice for the regressor variable in this model.

11-16 a) yˆ = 0.3299892 + 0.0041612( 95 x + 32)

yˆ = 0.3299892 + 0.0074902 x + 0.1331584

yˆ = 0.4631476 + 0.0074902 x

b) βˆ1 = 0.00749

11-17 Let x = engine displacement (cm3) and xold = engine displacement (in3)

a) The old regression equation is y = 39.2 − 0.0402xold

Because 1 in3 = 16.387 cm3, the new regression equation is
yˆ = 39.2 − 0.0402( x /16.387) = 39.2 − 0.0025 x

b) βˆ1 = −0.0025

11-18 βˆ0 + βˆ1 x = ( y − βˆ1 x ) + βˆ1 x = y

11-14
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-19 a) The slopes of both regression models will be the same, but the intercept will be shifted.
b) yˆ = 2132.41 − 36.9618 x

βˆ0 = 2625.39 βˆ0∗ = 2132.41

vs.
βˆ1 = −36.9618 βˆ1∗ = −36.9618

11-20 a) The least squares estimate minimizes ∑(y i − β xi ) 2 . Upon setting the derivative equal to zero, we

2∑ ( yi − β xi ) (− xi ) = 2[∑ − yi xi + β ∑ xi ] = 0
2
obtain

Therefore, βˆ =
∑yxi i
.
∑x
2
i

b) yˆ = 21.031461x . The model seems very appropriate—an even better fit.

45

40

35

30

25
chloride

20

15

10

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
watershed

11-15
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Section 11-4

βˆ0 − β 0 12.857
11-21 a) T0 = = = 12.4583
se( β 0 ) 1.032

P-value = 2[P( T8 > 12.4583)] and P-value < 2(0.0005) = 0.001

βˆ1 − β1 2.3445
T1 = = = 20.387
se( β1 ) 0.115

P-value = 2[P( T8 > 20.387)] and P-value < 2(0.0005) = 0.001

SS E 17.55
MS E = = = 2.1938
n−2 8
MS R 912.43
F0 = = = 415.913
MS E 2.1938

P-value is near zero

b) Because the P-value of the F-test ≈ 0 is less than α = 0.05, we reject the null hypothesis that
β1 = 0 at the 0.05 level of significance. This is the same result obtained from the T1 test. If the
assumptions are valid, a useful linear relationship exists.
c) σˆ 2 = MS E = 2.1938

βˆ0 − β 0 26.753
11-22 a) T0 = = = 11.2739
se( β 0 ) 2.373

P-value = 2[P( T14 > 11.2739)] and P-value < 2(0.0005) = 0.001

βˆ1 − β1 1.4756
T1 = = = 13.8815
se( β1 ) 0.1063

P-value = 2[P( T14 > 13.8815)] and P-value < 2(0.0005) = 0.001
Degrees of freedom of the residual error = 15 – 1 = 14.
Sum of squares regression = Sum of square Total – Sum of square residual error = 1500 – 94.8 =
1405.2
SS Regression 1405.2
MS Regression = = = 1405.2
1 1
MS R 1405.2
F0 = = = 192.4932
MS E 7.3

P-value is near zero

b) Because the P-value of the F-test ≈ 0 is less than α = 0.05, we reject the null hypothesis that
β1 = 0 at the 0.05 level of significance. This is the same result obtained from the T1 test. If the
assumptions are valid, a useful linear relationship exists.
c) σˆ 2 = MS E = 7.3

11-16
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-23 a) 1) The parameter of interest is the regressor variable coefficient, β1

2) H 0 : β1 = 0

3) H1 : β1 ≠ 0

4) α = 0.01
5) The test statistic is
MS R SS R /1
f0 = =
MS E SS E /(n − 2)

6) Reject H0 if f0 > fα,1,12 where f0.01,1,12 = 9.33

7) Using results from Exercise 11-1

SS R = βˆ1 S xy = −2.3298017(−59.057143)
= 137.59
SS E = S yy − SS R
= 159.71429 − 137.59143
= 22.123
137.59
f0 = = 74.63
22.123 /12
8) Since 74.63 > 9.33 reject H0 and conclude that compressive strength is significant in predicting
intrinsic permeability of concrete at α = 0.01. We can therefore conclude that the model specifies a
useful linear relationship between these two variables.
P-value ≅ 0.000002

SS E 22.123 σˆ 2 1.8436
b) σˆ 2 = MS E = = = 1.8436 and se( βˆ1 ) = = = 0.2696
n−2 12 S xx 25.3486

⎡1 x2 ⎤ ⎡ 1 3.07142 ⎤
c) se( βˆ0 ) = σˆ 2 ⎢ + ⎥ = 1.8436 ⎢ + ⎥ = 0.9043
⎣ n S xx ⎦ ⎣14 25.3486 ⎦
11-24 a) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 : β1 = 0

3) H1 : β1 ≠ 0

4) α = 0.01
5) The test statistic is
MS R SS R /1
f0 = =
MS E SS E /(n − 2)

6) Reject H0 if f0 > fα,1,18 where f0.01,1,18 = 8.29

7) Using the results from Exercise 11-2

11-17
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

SS R = βˆ1 S xy = (0.0041612)(141.445)
= 0.5886
SS E = S yy − SS R
= (8.86 − 12.75 ) − 0.5886
2

20

= 0.143275
0.5886
f0 = = 73.95
0.143275 /18
8) Since 73.95 > 8.29, reject H0 and conclude the model specifies a useful relationship at α = 0.01.
P-value ≅ 0.000001

σˆ 2 .00796
b) se( βˆ1 ) = = = 4.8391× 10−4
S xx 33991.6

⎡1 x ⎤ ⎡1 73.92 ⎤
se( βˆ0 ) = σˆ 2 ⎢ + ⎥ = .00796 ⎢ + ⎥ = 0.04091
⎣ n S xx ⎦ ⎣ 20 33991.6 ⎦

11-25 a)
Regression Analysis: Rating Pts versus Yds per Att

The regression equation is

Rating Pts = 14.2 + 10.1 Yds per Att

Predictor Coef SE Coef T P

Constant 14.195 9.059 1.57 0.128
Yds per Att 10.092 1.288 7.84 0.000

S = 5.21874 R-Sq = 67.2% R-Sq(adj) = 66.1%

Analysis of Variance
Source DF SS MS F P
Regression 1 1672.5 1672.5 61.41 0.000
Residual Error 30 817.1 27.2
Total 31 2489.5

Refer to the ANOVA

H 0 : β1 = 0
H1 : β1 ≠ 0
α = 0.01
Because the P-value = 0.000 < α = 0.01, reject H0. If the assumptions are valid, we conclude that there
is a useful linear relationship between these two variables.

b) σˆ 2 = 27.2

11-18
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

σˆ 2 27.2
se( βˆ1 ) = = = 1.287
S xx 16.422

⎡1 x2 ⎤ ⎡1 72 ⎤
se( βˆ0 ) = σˆ 2 ⎢ + ⎥ = 27.2 ⎢ + ⎥ = 9.056
⎣ n S xx ⎦ ⎣ 32 16.422 ⎦

c) 1) The parameter of interest is the regressor variable coefficient β1.

2) H 0 : β1 = 10

3) H1 : β1 ≠ 10

4) α = 0.01

βˆ1 − β1,0
5) The test statistic is t0 =
se( βˆ1 )

6) Reject H0 if t0 < −tα/2,n-2 where −t0.005,30 = −2.750 or t0 > t0.005,30 = 2.750

7) Using the results from Exercise 10-6
10.092 − 10
t0 = = 0.0715
1.287
8) Because 0.0715 < 2.750, fail to reject H0. There is not enough evidence to conclude that the slope
differs from 10 at α = 0.01.

11-26 Refer to ANOVA for Exercise 11-4

a) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 : β1 = 0

3) H1 : β1 ≠ 0

4) α = 0.05, using t-test

βˆ1
5) The test statistic is t0 =
se( βˆ1 )

6) Reject H0 if t0 < −tα/2,n-2 where −t0.025,22 = −2.074 or t0 > t0.025,22 = 2.074

7) Using the results from Exercise 11-5
3.32437
t0 = = 8.518
0.390276
8) Since 8.518 > 2.074 reject H0 and conclude the model is useful α = 0.05.

b) 1) The parameter of interest is the slope, β1

2) H 0 : β1 = 0

3) H1 : β1 ≠ 0

4) α = 0.05
MSR SSR /1
5) The test statistic is f 0 = =
MSE SSE /(n − 2)

11-19
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

6) Reject H0 if f0 > fα,1,22 where f0.01,1,22 = 4.303

7) Using the results from Exercise 10-5
636.15569 /1
f0 = = 72.5563
192.89056 / 22
8) Since 72.5563 > 4.303, reject H0 and conclude the model is useful at a significance α = 0.05.
The F-statistic is the square of the t-statistic. The F-test is a restricted to a two-sided test, whereas
the t-test could be used for one-sided alternative hypotheses.

σˆ 2 8.7675
c) se( βˆ1 ) = = = .39027
S xx 57.5631

⎡1 x ⎤ ⎡ 1 6.4049 2 ⎤
se( βˆ0 ) = σˆ 2 ⎢ + ⎥ = 8.7675 ⎢ + ⎥ = 2.5717
⎣ n S xx ⎦ ⎣ 24 57.5631 ⎦

d) 1) The parameter of interest is the intercept, β0.

2) H 0 : β1 = 0

3) H1 : β1 ≠ 0

4) α = 0.05, using t-test

βˆ0
5) The test statistic is t0 =
se( βˆ0 )

6) Reject H0 if t0 < −tα/2,n-2 where −t0.025,22 = −2.074 or t0 > t0.025,22 = 2.074

7) Using the results from Exercise 11-5
13.3201
t0 = = 5.179
2.5717
8) Since 5.179 > 2.074 reject H0 and conclude the intercept is not zero at α = 0.05.

11-27 Refer to the ANOVA for Exercise 10-65

a) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 : β1 = 0

3) H1 : β1 ≠ 0

4) α = 0.05
MS R SS R /1
5) The test statistic is f 0 = =
MS E SS E /(n − 2)

6) Reject H0 if f0 > fα,1,10 where f0.05,1,10 = 10.04

7) Using the results from Exercise 10-6
280583.12 /1
f0 = = 74334.4
37.746089 /10
8) Since 74334.4 > 10.04, reject H0 and conclude the model is useful α = 0.05. P-value < 0.000001

b) se( β̂1 ) = 0.0337744, se( βˆ0 ) = 1.66765

11-20
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

c) 1) The parameter of interest is the regressor variable coefficient, β1.

2) H 0 : β1 = 10

3) H1 : β1 ≠ 10

4) α = 0.05, α/2 = 0.025

βˆ1 − β1,0
5) The test statistic is t0 =
se( βˆ1 )

6) Reject H0 if t0 < −tα/2,n-2 where −t0.025,10 = −2.228 or t0 > t0.025,10 = 2.228

7) Using the results from Exercise 10-6
9.21 − 10
t0 = = −23.37
0.0338
8) Since −23.37 < −2.228 reject H0 and conclude the slope is not 10 at α = 0.05. P-value ≈ 0.
d) H0: β0 = 0 H1: β0 ≠ 0
−6.3355 − 0
t0 = = −3.8
1.66765

P-value < 0.005. Reject H0 and conclude that the intercept should be included in the model.

11-28 Refer to the ANOVA for Exercise 11-6

(a) H 0 : β1 = 0; H1 : β1 ≠ 0

MS R 385.18
f0 = = = 27.49
MS E 14.01
F0.01,1,19 = 8.18

Reject the null hypothesis and conclude that the slope is not zero. The exact P-value is P = 0.0000463

(b) From the computer output from Exercise 11-6:

se( βˆ0 ) = 2.006, se( βˆ1 ) = 0.007671

(c) H 0 : β1 = −0.05; H1 : β1 < −0.05

βˆ1 − βˆ1,0 −0.040216 − (−0.05) 0.090216

t0 = = = = 11.76
se( βˆ1 ) 0.007671 0.007671
t0.01,19 = 2.539, since t0 is not less than − t0.01,19 = −2.539, do not reject H 0
P ≅ 1.0

(d) H 0 : β 0 = 0; H1 : β 0 ≠ 0

βˆ0 − βˆ0,0 39.156 − 0

t0 = = = 19.52
se( βˆ0 ) 2.006
t0.005,19 = 2.861, since |t0 | >t0.005,19 reject H 0
P = 4.95E − 14 ≅ 0

11-21
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-29 Refer to the ANOVA for Exercise 11-7

a) H 0 : β1 = 0

H1 : β1 ≠ 0

α = 0.05
f 0 = 44.0279
f.05,1,11 = 4.84
f 0 > fα ,1,11

Therefore, reject H0. P-value = 0.00004.

b) se( βˆ1 ) = 0.0104524

se( βˆ0 ) = 9.84346

c) H 0 : β 0 = 0

H1 : β 0 ≠ 0

α = 0.05
t0 = −1.67718
t.025,11 = 2.201
| t0 |</ −tα / 2,11

Therefore, do not reject H0. P-value = 0.12166.

11-30 Refer to the ANOVA for Exercise 11-8

a) H 0 : β1 = 0

H1 : β1 ≠ 0

α = 0.05
f 0 = 53.50
f.05,1,18 = 4.414
f 0 > fα ,1,18

Therefore, reject H0. P-value = 0.0000009.

b) se( βˆ1 ) = 0.0256613

se( βˆ0 ) = 2.13526

c) H 0 : β 0 = 0

H1 : β 0 ≠ 0

α = 0.05

11-22
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

t0 = − 5.079
t.025,18 = 2.101
| t0 |> tα / 2,18

Therefore, reject H0. P-value = 0.000078.

11-31 Refer to ANOVA for Exercise 11-11

a) H 0 : β1 = 0

H1 : β1 ≠ 0

α = 0.01
f 0 = 156.67
f.01,1,18 = 8.285
f 0 > fα ,1,18

Therefore, reject H0. P-value < 0.00001.

b) se( βˆ1 ) = 310.8

se( βˆ0 ) = 20.34

c) H 0 : β1 = −206.84

H1 : β1 ≠ −206.84

α = 0.01
−254.55 − ( −206.84)
t0 = = −2.3456
20.34
t.005,18 = 2.878
| t0 |>/ −tα / 2,18

Therefore, do not reject H0. P-value = 0.0153(2) = 0.0306.

d) H 0 : β 0 = 0

H1 : β 0 ≠ 0

α = 0.01
t0 = 58.20
t.005,18 = 2.878

t0 > tα / 2,18 , therefore, reject H0. P-value < 0.00001.

e) H 0 : β 0 = 17236.89

H1 : β 0 > 17236.89

α = 0.01

11-23
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

18090.2 − 17236.89
t0 = = 2.7455
310.8
t.01,18 = 2.552

t0 > tα ,18 , therefore reject H0. P-value = 0.0064.

11-32 Refer to ANOVA for Exercise 11-10

a) H 0 : β1 = 0

H1 : β1 ≠ 0

α = 0.05
f 0 = 92.224
f 0.05,1,16 = 4.49
f 0 > fα ,1,16

Therefore, reject H0.

b) P-value < 0.00001

c) se( βˆ1 ) = 2.14169

se( βˆ0 ) = 1.93591

d) H 0 : β 0 = 0

H1 : β 0 ≠ 0

α = 0.05
t0 = 0.243
t0.025,16 = 2.12
t0 >/ tα / 2,16

Therefore, do not reject H0. There is not sufficient evidence to conclude that the intercept differs
from zero. Based on this test result, the intercept could be removed from the model.

11-33 a) Refer to ANOVA for Exercise 11-13.

H 0 : β1 = 0
H1 : β1 ≠ 0
α = 0.01
Because the P-value = 0.000 < α = 0.01, reject H0. There is evidence of a linear relationship between
these two variables.

b) σˆ 2 = 0.083
The standard errors for the parameters can be obtained from the computer output or calculated as
follows.

11-24
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

σˆ 2 0.083
se( βˆ1 ) = = = 0.014
S xx 420.91

⎡1 x2 ⎤ ⎡ 1 10.092 ⎤
se( βˆ0 ) = σˆ 2 ⎢ + ⎥ = 0.083 ⎢ + ⎥ = 0.1657
⎣ n S xx ⎦ ⎣11 420.91 ⎦

c) 1) The parameter of interest is the intercept β0.

2) H 0 : β 0 = 0

3) H1 : β 0 ≠ 0

4) α = 0.01
β0
5) The test statistic is t0 =
se( β 0 )

6) Reject H0 if t0 < −tα / 2, n − 2 where −t0.005,9 = −3.250 or t0 > tα / 2,n − 2 where t0.005,9 = 3.250

7) Using the results from Exercise 11-13

0.6578
t0 = = 3.97
0.1657
8) Because t0 = 3.97 > 3.250 reject H0 and conclude the intercept is not zero at α = 0.01.

11-34 a) Refer to ANOVA for Exercise 11-14

H 0 : β1 = 0
H1 : β1 ≠ 0
α = 0.01
Because the P-value = 0.000 < α = 0.01, reject H0. There is evidence of a linear relationship between
these two variables.
b) Yes

c) σˆ 2 = 1.118

σˆ 2 1.118
se( βˆ1 ) = = = 0.0436
S xx 588

⎡1 x2 ⎤ ⎡ 1 65.67 2 ⎤
se( βˆ0 ) = σˆ 2 ⎢ + ⎥ = 1.118 ⎢ + ⎥ = 2.885
⎣ n S xx ⎦ ⎣9 588 ⎦

11-35 a) H 0 : β1 = 0

H1 : β1 ≠ 0
α = 0.01
Because the P-value = 0.310 > α = 0.01, fail to reject H0. There is not sufficient evidence of a linear
relationship between these two variables.

The regression equation is

11-25
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

BMI = 13.8 + 0.256 Age

Predictor Coef SE Coef T P

Constant 13.820 9.141 1.51 0.174
Age 0.2558 0.2340 1.09 0.310
S = 5.53982 R-Sq = 14.6% R-Sq(adj) = 2.4%

Analysis of Variance
Source DF SS MS F P
Regression 1 36.68 36.68 1.20 0.310
Residual Error 7 214.83 30.69
Total 8 251.51

b) σˆ 2 = 30.69, se( βˆ1 ) = 0.2340, se( βˆ0 ) = 9.141 from the computer output

⎡1 x2 ⎤ ⎡ 1 38.2562 ⎤
c) se( βˆ0 ) = σˆ 2 ⎢ + ⎥ = 30.69 ⎢ + ⎥ = 9.141
⎣ n S xx ⎦ ⎣ 9 560.342 ⎦

βˆ1 b
After the transformation βˆ1∗ = βˆ1 , S xx
∗
= a 2 S xx , x ∗ = ax , βˆ0∗ = bβˆ0 , and
11-36 t0 =
σˆ / S xx
2 a

σˆ ∗ = bσˆ . Therefore, t0∗ = b βˆ1 / a

= t0 .
(bσˆ ) 2 / a 2 S xx

|10 − (12.5) |
11-37 d= = 0.331
5.5 31/16.422
Assume α = 0.05, from Chart VII (e) and interpolating between the curves for n = 30 and n = 40,
β ≅ 0.55

βˆ
11-38 a) has a t distribution with n − 1 degree of freedom.
σˆ 2
∑x i
2

b) From Exercise 11-17, βˆ = 21.031461, σˆ = 3.611768, and ∑x 2

i = 14.7073 .

The t-statistic in part (a) is 22.3314 and H 0 : β 0 = 0 is rejected at usual α values.

11-26
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Sections 11-5 and 11-6

11-39 tα/2,n-2 = t0.025,12 = 2.179

a) 95% confidence interval on β1 .

βˆ1 ± tα / 2, n − 2 se( βˆ1 )

−2.3298 ± t.025,12 (0.2696)
−2.3298 ± 2.179(0.2696)
−2.9173. ≤ β1 ≤ −1.7423.

b) 95% confidence interval on β 0 .

βˆ0 ± t.025,12 se( βˆ0 )

48.0130 ± 2.179(0.5959)
46.7145 ≤ β 0 ≤ 49.3115.

c) 95% confidence interval on μ when x 0 = 2.5 .

μˆY | x0 = 48.0130 − 2.3298(2.5) = 42.1885

( x0 − x ) 2
μˆY | x0 ± t.025,12 σˆ 2 ( 1n + S xx
)

(2.5 −3.0714)2
42.1885 ± (2.179) 1.844( 141 + 25.3486
)
42.1885 ± 2.179(0.3943)
41.3293 ≤ μˆY | x0 ≤ 43.0477

d) 99% on prediction interval when x 0 = 2.5 .

(x −x ) 2
yˆ 0 ± t0.005,12 σˆ 2 (1 + 1n + 0S )
xx

(2.5 −3.0714) 2
42.1885 ± 3.055 1.844(1 + 14
1
+ 25.348571
)
42.1885 ± 3.055(1.4056)
37.8944 ≤ y0 ≤ 46.4826

It is wider because it depends on both the errors associated with the fitted model and the
future observation.

11-40 tα/2,n-2 = t0.005,18 = 2.878

a) βˆ1 ± ( t0.005,18 ) se( βˆ1 )

0.0041612 ± (2.878)(0.000484)
0.0027682 ≤ β1 ≤ 0.0055542

b) βˆ0 ± ( t0.005,18 ) se( βˆ0 )

0.3299892 ± (2.878)(0.04095)
0.212250 ≤ β 0 ≤ 0.447728

c) 99% confidence interval on μ when x0 = 85D F .

11-27
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

μˆY | x0 = 0.683689
( x0 − x ) 2
μˆY | x0 ± t.005,18 σˆ 2 ( 1n + S xx
)

(85 − 73.9)2
0.683689 ± (2.878) 0.00796( 20
1
+ 33991.6
)
0.683689 ± 0.0594607
0.6242283 ≤ μˆY | x0 ≤ 0.7431497

d) 99% prediction interval when x0 = 90°F .

yˆ 0 = 0.7044949
(x −x) 2
yˆ 0 ± t.005,18 σˆ 2 (1 + 1n + 0S )
xx

(90 − 73.9) 2
0.7044949 ± 2.878 0.00796(1 + 20
1
+ 33991.6
)
0.7044949 ± 0.263567
0.420122 ≤ y0 ≤ 0.947256

11-41 tα / 2,n − 2 = t0.025,30 = 2.042

a) 99% confidence interval on β1

βˆ1 ± tα / 2,n − 2 se( βˆ1 )

10.092 ± t0.005,30 (1.287)
10.092 ± 2.750(1.287)
6.553 ≤ β1 ≤ 13.631

b) 99% confidence interval on β 0

βˆ0 ± tα / 2,n − 2 se( βˆ0 )

14.195 ± t0.005,30 (9.056)
14.195 ± 2.75(9.056)
−10.709 ≤ βˆ ≤ 39.099
0

c) 99% confidence interval for the mean rating when the average yards per attempt is 8.0
μˆ = 14.195 + 10.092(8.0) = 94.931

⎛ ( x0 − x )2 ⎞
2 1
μ ± t0.005,30 σ
ˆ ˆ ⎜ + ⎟
⎜n S xx ⎟
⎝ ⎠
⎛ 1 ( 8 − 7 )2 ⎞
94.931 ± 2.75 27.2 ⎜ + ⎟
⎜ 32 16.422 ⎟
⎝ ⎠
90.577 ≤ μ ≤ 99.285

d) 99% prediction interval on x0 = 8.0

11-28
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

⎛ 1 ( x − x )2 ⎞
yˆ ± t0.005,30 σˆ 2 ⎜ 1 + + 0 ⎟
⎜ n S xx ⎟
⎝ ⎠
⎛ 1 (8 − 7 ) ⎞
2
94.931 ± 2.75 27.2 ⎜ 1 + + ⎟
⎜ 32 16.422 ⎟
⎝ ⎠
79.943 ≤ μ ≤ 109.919

11-42
Regression Analysis: Price versus Taxes

The regression equation is

Price = 13.3 + 3.32 Taxes

Predictor Coef SE Coef T P

Constant 13.320 2.572 5.18 0.000
Taxes 3.3244 0.3903 8.52 0.000

S = 2.96104 R-Sq = 76.7% R-Sq(adj) = 75.7%

Analysis of Variance

Source DF SS MS F P
Regression 1 636.16 636.16 72.56 0.000
Residual Error 22 192.89 8.77
Total 23 829.05

a) 3.32437 – 2.074(0.3903) = 2.515 ≤ β1 ≤ 3.32437 + 2.074(0.3903) = 4.134

b) 13.320 – 2.074(0.3903) = 7.985 ≤ β0 ≤ 13.320 + 2.074(0.39028) = 18.655

(7.5 − 6.40492)2
c) 38.253 ± (2.074) 8.76775( 24
1
+ 57.563139
)

38.253 ± 1.5353
36.7177 ≤ μˆY | x0 ≤ 39.7883

(7.5− 6.40492) 2
d) 38.253 ± (2.074) 8.76775(1 + 24
1
+ 57.563139
)

38.253 ± 6.3302
31.9228 ≤ y0 ≤ 44.5832

11-43
Regression Analysis: Usage versus Temperature

The regression equation is

11-29
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Usage = 130 + 7.59 Temperature

Predictor Coef SE Coef T P

Constant 129.974 0.707 183.80 0.000
Temperature 7.59262 0.05798 130.95 0.000

S = 1.83431 R-Sq = 99.9% R-Sq(adj) = 99.9%

Analysis of Variance

Source DF SS MS F P
Regression 1 57701 57701 17148.85 0.000
Residual Error 10 34 3
Total 11 57734

a) 7.59262 – 2.228(0.05798) = 7.463 ≤ β1 ≤ 7.59262 + 2.228(0.05798) = 7.721

b) 129.974 – 2.228(0.707) = 128.399 ≤ β0 ≤ 129.974 + 2.228(0.707) = 131.549

(13−8.08)2
c) 228.67 ± (2.228) 3( 121 + 1000.917
)

228.67 ± 1.26536
227.4046 ≤ μˆ Y|x 0 ≤ 229.9354

(13−8.08) 2
d) 228.67 ± (2.228) 3(1 + 12
1
+ 1000.917
)

228.67 ± 4.061644
224.6084 ≤ y0 ≤ 232.73164

It is wider because the prediction interval includes errors for both the fitted model and for a future
observation.

11-44 Refer to the ANOVA for Exercise 11-6.

(a) t0.025, 19 = 2.093
34.96 ≤ β 0 ≤ 43.36; −0.0563 ≤ β1 ≤ −0.0241

(b) Descriptive Statistics: x = displacement

Sum of
Variable n Mean Sum Squares
x 21 238.9 5017.0 1436737.0

11-30
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

yˆ = 33.15 when x = 150

⎡ 1 (150 − 238.9) 2 ⎤
33.15 ± 2.093 14.01 ⎢ + ⎥
⎣⎢ 21 1, 436, 737.0 ⎦⎥
33.15 ± 1.8056
31.34 ≤ μY | x =150 ≤ 34.96

(c) yˆ = 33.15 when x = 150

⎡ 1 (150 − 238.9) 2 ⎤
33.15 ± 2.093 14.01 ⎢1 + + ⎥
⎢⎣ 21 1, 436, 737.0 ⎥⎦
33.15 ± 8.0394
25.11 ≤ Y0 ≤ 41.19

11-45 a) 0.03689 ≤ β1 ≤ 0.10183

b) −47.0877 ≤ β 0 ≤ 14.0691

( 910 − 939 ) 2
c) 46 . 6041 ± ( 3 . 106 ) 7 . 324951 ( 131 + 67045 . 97
)

46 .6041 ± 2 .514401
44 .0897 ≤ μ y | x 0 ≤ 49 .1185 )

(910 −939) 2
d) 46.6041 ± (3.106) 7.324951(1 + 13
1
+ 67045.97
)

46.6041 ± 8.779266
37.8298 ≤ y0 ≤ 55.3784

11-46 a) 0.11756 ≤ β1 ≤ 0.22541

b) −14.3002 ≤ β 0 ≤ −5.32598

(85 −82.3) 2
c) 4.76301 ± (2.101) 1.982231( 20
1
+ 3010.2111
)

4.76301 ± 0.6772655
4.0857 ≤ μ y | x0 ≤ 5.4403

(85−82.3)2
d) 4.76301 ± (2.101) 1.982231(1 + 20
1
+ 3010.2111
)

4.76301 ± 3.0345765
1.7284 ≤ y0 ≤ 7.7976

11-47 a) 201.552≤ β1 ≤ 266.590

b) − 4.67015≤ β0 ≤ −2.34696

(30 − 24.5)2
c) 128.814 ± (2.365) 398.2804( 19 + 1651.4214
)

11-31
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

128.814 ± 16.980124
111.8339 ≤ μ y| x0 ≤ 145.7941

11-48 a) 14.3107 ≤ β1 ≤ 26.8239

b) −5.18501 ≤ β 0 ≤ 6.12594

(1− 0 .806111 ) 2
c) 21 .038 ± ( 2 .921 ) 13 .8092 ( 181 + 3 .01062 )

21.038 ± 2.8314277
18.2066 ≤ μ y| x0 ≤ 23.8694

(1− 0 .806111 ) 2
d) 21 .038 ± ( 2.921) 13 .8092 (1 + 181 + 3.01062 )

21 .038 ± 11 .217861
9 .8201 ≤ y 0 ≤ 32 .2559

11-49 a) −313.0885 ≤ β1 ≤ −196.0115

b) 17195.7176 ≤ β 0 ≤ 18984.6824

(20 −13.3375) 2
c) 12999.2 ± 2.878 460992( 20
1
+ 1114.6618
)

12999.2 ± 585.64
12413.56 ≤ μ y| x0 ≤ 13584.84

(20 −13.3375)2
d) 12999.2 ± 2.878 460992(1 + 20
1
+ 1114.6618
)

12999.2 ± 2039.93
10959.27 ≤ y0 ≤ 15039.14

11-50 tα / 2,n − 2 = t0.005,5 = 4.032

a) 99% confidence interval on β̂1

βˆ1 ± tα / 2,n − 2 se( βˆ1 )

−0.034 ± t0.001,5 (0.026)
−0.034 ± 4.032(0.026)
−0.1388 ≤ βˆ ≤ 0.0708
1

b) 99% confidence interval on β 0

11-32
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

βˆ0 ± tα / 2,n − 2 se( βˆ0 )

55.63 ± t0.005,5 (32.11)
55.63 ± 4.032(32.11)
−73.86 ≤ βˆ ≤ 185.12
0

c) 99% confidence interval for the mean length when x = 1500:

μˆ = 55.63 − 0.034(1500) = 4.63

⎛1 ( x0 − x )2 ⎞⎟
μˆ ± t0.005,5 σˆ 2 ⎜ +
⎜n S xx ⎟
⎝ ⎠
⎛ 1 (1500 − 1242.86 )2 ⎞
4.63 ± 4.032 77.33 ⎜ + ⎟
⎜7 117142.8 ⎟
⎝ ⎠
4.63 ± 4.032(7.396)
−25.19 ≤ μ ≤ 34.45

d) 99% prediction interval when x0 = 1500

⎛ 1 ( x − x )2 ⎞
yˆ ± t0.005,5 σˆ 2 ⎜ 1 + + 0 ⎟
⎜ n S xx ⎟
⎝ ⎠
⎛ 1 (1500 − 1242.86 )2 ⎞
4.63 ± 4.032 77.33 ⎜ 1 + + ⎟
⎜ 7 117142.8 ⎟
⎝ ⎠
4.63 ± 4.032(11.49)
−41.7 ≤ y0 ≤ 50.96

It’s wider because it depends on both the error associated with the fitted model as well as that of the
future observation.

11-51 Refer to the Minitab output in Exercise 11-13

tα / 2,n − 2 = t0.005,9 = 3.250

a) 99% confidence interval for β̂1

βˆ1 ± tα / 2,n − 2 se( βˆ1 )

0.178 ± t0.005,9 (0.014)
0.178 ± 3.250(0.014)
0.1325 ≤ βˆ ≤ 0.2235
1

b) 99% confidence interval on β 0

βˆ0 ± tα / 2,n − 2 se( βˆ0 )

0.6578 ± t0.005,9 (0.1657)
0.6578 ± 3.250(0.1657)
0.119 ≤ βˆ ≤ 1.196
0

c) 95% confidence interval on μ when x0 = 8

11-33
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

μˆ y| x0 = 0.658 + 0.178(8) = 2.082

⎛1 ( x0 − x )2 ⎞⎟
μˆ y| x0 ± t0.025,9 σˆ 2 ⎜ +
⎜n S xx ⎟
⎝ ⎠
⎛ 1 ( 8 − 10.09 )2 ⎞
2.082 ± 2.262 0.083 ⎜ + ⎟
⎜ 11 420.91 ⎟
⎝ ⎠
1.87 ≤ μ y| x0 ≤ 2.29

11-34
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Section 11-7

S 2 25.35
11-52 R 2 = βˆ12 XX = ( −2.330 ) = 0.8617
SYY 159.71

The model accounts for 86.17% of the variability in the data.

11-53 Refer to the Minitab output in Exercise 11-3.

a) R 2 = 0.672
The model accounts for 67.2% of the variability in the data.

b) There is no major departure from the normality assumption in the following graph.

c) The assumption of constant variance appears reasonable.

11-35
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-54 Use the results from Exercise 11-5 to answer the following questions.
a) SalePrice Taxes Predicted Residuals
25.9 4.9176 29.6681073 -3.76810726
29.5 5.0208 30.0111824 -0.51118237
27.9 4.5429 28.4224654 -0.52246536
25.9 4.5573 28.4703363 -2.57033630
29.9 5.0597 30.1405004 -0.24050041
29.9 3.8910 26.2553078 3.64469225
30.9 5.8980 32.9273208 -2.02732082
28.9 5.6039 31.9496232 -3.04962324
35.9 5.8282 32.6952797 3.20472030
31.5 5.3003 30.9403441 0.55965587
31.0 6.2712 34.1679762 -3.16797616
30.9 5.9592 33.1307723 -2.23077234
30.0 5.0500 30.1082540 -0.10825401
36.9 8.2464 40.7342742 -3.83427422
41.9 6.6969 35.5831610 6.31683901
40.5 7.7841 39.1974174 1.30258260
43.9 9.0384 43.3671762 0.53282376
37.5 5.9894 33.2311683 4.26883165
37.9 7.5422 38.3932520 -0.49325200
44.5 8.7951 42.5583567 1.94164328
37.9 6.0831 33.5426619 4.35733807
38.9 8.3607 41.1142499 -2.21424985
36.9 8.1400 40.3805611 -3.48056112
45.8 9.1416 43.7102513 2.08974865

b) Assumption of normality does not seem to be violated since the data appear to fall along a straight
line.

Normal Probability Plot

99.9

99

95
cumulative percent

80

50

20

0.1

-4 -2 0 2 4 6 8

Residuals

11-36
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

c) There are no serious departures from the assumption of constant variance. This is evident by the
random pattern of the residuals.
Plot of Residuals versus Predicted Plot of Residuals versus Taxes

8 8

6 6

4 4

Residuals

Residuals
2 2

0 0

-2 -2

-4 -4

26 29 32 35 38 41 44 3.8 4.8 5.8 6.8 7.8 8.8 9.8

Predicted Values Taxes

d) R 2 ≡ 76.73% ;

11-55 Use the results of Exercise 11-5 to answer the following questions

a) R 2 = 99.986%; The proportion of variability explained by the model.

b) Yes, normality seems to be satisfied because the data appear to fall along the straight line.

Normal Probability Plot

99.9

99

95
cumulative percent

80

50

20

0.1

-2.6 -0.6 1.4 3.4 5.4

Residuals

c) There might be lower variance at the middle settings of x. However, this data does not indicate a
serious departure from the assumptions.

Plot of Residuals versus Predicted Plot of Residuals versus Temperature

5.4 5.4

3.4 3.4
Residuals
Residuals

1.4 1.4

-0.6 -0.6

-2.6 -2.6

180 280 380 480 580 680 21 31 41 51 61 71 81

Predicted Values Temperature

11-37
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-56 a) R 2 = 20.1121%
b) These plots might indicate the presence of outliers, but no real problem with assumptions.

c) The normality assumption appears marginal.

Normal Probability Plot of the Residuals

(response is y)

10
Residual

-10
-2 -1 0 1 2

Normal Score

11-57 a) R 2 = 0.879397
b) No departures from constant variance are noted.
Residuals Versus x Residuals Versus the Fitted Values
(response is y) (response is y)

30 30

20 20

10 10
Residual

Residual

0 0

-10 -10

-20 -20

-30 -30
0 10 20 30 40 80 130 180 230

x Fitted Value

c) Normality assumption appears reasonable.

Normal Probability Plot of the Residuals

(response is y)

30

20

10
Residual

-10

-20

-30
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Normal Score

11-38
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-58 a) R 2 = 71.27%
b) No major departure from normality assumptions.
Normal Probability Plot of the Residuals
(response is y)

Residual
0

-1

-2
-2 -1 0 1 2

Normal Score

c) Assumption of constant variance appears reasonable.

Residuals Versus x
(response is y)
Residuals Versus the Fitted Values
(response is y)

3
3

2
2

1
1
Residual

Residual
0 0

-1 -1

-2 -2
60 70 80 90 100 0 1 2 3 4 5 6 7 8

x Fitted Value

11-59 a) R 2 = 85 . 22 %
b) Assumptions appear reasonable, but there is a suggestion that variability increases slightly with ŷ .

Residuals Versus x
Residuals Versus the Fitted Values
(response is y)
(response is y)

5
5
Residual

Residual

0 0

-5 -5
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0 10 20 30 40

x Fitted Value

c) Normality assumption may be questionable. There is some “bending” away from a line in the tails
of the normal probability plot.

11-39
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Normal Probability Plot of the Residuals

(response is y)

Residual
0

-5
-2 -1 0 1 2

Normal Score

11-60 a)
The regression equation is
Compressive Strength = - 2150 + 185 Density

Predictor Coef SE Coef T P

Constant -2149.6 332.5 -6.46 0.000
Density 184.55 11.79 15.66 0.000

S = 339.219 R-Sq = 86.0% R-Sq(adj) = 85.6%

Analysis of Variance

Source DF SS MS F P
Regression 1 28209679 28209679 245.15 0.000
Residual Error 40 4602769 115069
Total 41 32812448

b) Because the P-value = 0.000 < α = 0.01, the model is significant.

c) σˆ 2 = 115069
SS R SS 28209679
d) R 2 = = 1− E = = 0.8597 = 85.97%
SST SST 32812448

The model accounts for 85.97% of the variability in the data.

e)

No major departure from the normality assumption.

f)

11-40
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Assumption of constant variance appears reasonable.

11-61 a) R 2 = 0.896947 89% of the variability is explained by the model.

b) Yes, the two points with residuals much larger in magnitude than the others seem unusual.

2
c) Rnew model = 0.8799

Smaller, because the older model is better able to account for the variability in the data with these
two outlying data points removed.

d) σˆ old
2
model = 460992

σˆ new
2
model = 188474

Yes, reduced more than 50%, because the two removed points accounted for a large amount of the
error.

11-62 a)

11-41
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

yˆ = 0.55 + 0.05237 x

b) H0 : β1 = 0 H1 : β1 ≠ 0 α = 0.05

f0 = 7.41
f.05,1,12 = 4.75
f0 > fα ,1,12

Reject H0.

c) σˆ 2 = 26.97

d) σˆ orig 2 = 7.502

The new estimate is larger because the new point added additional variance that was not accounted
for by the model.
e) yˆ = 0.55 + 0.05237(860) = 45.5882

e = y − yˆ = 60 − 45.5882 = 14.4118

Yes, e14 is especially large compared to the other residuals.

f) The one added point is an outlier and the normality assumption is not as valid with the point
included.

11-42
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

g) Constant variance assumption appears valid except for the added point.

11-43
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-63 Yes, when the residuals are standardized the unusual residuals are easier to identify.
1.11907 −0.75653 −0.13113 0.68314 −2.49705 −2.26424 0.51810
0.48210 0.11676 0.40780 0.22274 −0.93513 0.88167 0.76461
−0.49995 0.99241 0.12989 0.39831 1.15898 −0.82134

11-64 For two random variables X1 and X2,

V ( X 1 + X 2 ) = V ( X 1 ) + V ( X 2 ) + 2Cov( X 1 , X 2 )

Then,

V (Yi − Yˆi ) = V (Yi ) + V (Yˆi ) − 2Cov(Yi , Yˆi )

= σ 2 + V ( βˆ0 + βˆ1 xi ) − 2σ 2 ⎡ 1n + iS xx ⎤
(x −x) 2

⎣ ⎦
2 ⎡1 ( xi − x ) 2 ⎤ 2 ⎡1 ( xi − x )2 ⎤
= σ + σ n + S xx − 2σ n + S xx
2
⎣ ⎦ ⎣ ⎦
= σ 1 − ( 1n + iS xx ) ⎤
2 ⎡ ( x − x ) 2

⎣ ⎦
a) Because ei is divided by an estimate of its standard error (when σ 2 is estimated by σˆ 2 ), ri has
approximately unit variance.
b) No, the term in brackets in the denominator is necessary.
c) If xi is near x and n is reasonably large, ri is approximately equal to the standardized residual.
d) If xi is far from x , the standard error of ei is small. Consequently, extreme points are better fit by
least squares regression than points near the middle range of x. Because the studentized residual at
any point has variance of approximately one, the studentized residuals can be used to compare the fit
of points to the regression line over the range of x.

( n − 2)(1 − SSSyyE ) S yy − SS E S yy − SS E
Using R 2 = 1 − , F0 = = =
SS E
11-65 S yy SS E
S yy
SS E
n−2
σˆ 2

Also,
SS E = ∑ ( yi − βˆ 0 − βˆ1 xi ) 2
= ∑ ( yi − y − βˆ1 ( xi − x )) 2
= ∑ ( yi − y ) + βˆ1 ∑ ( x − x ) − 2βˆ ∑ ( yi − y )(xi − x )
2 2
i 1

= ∑ ( yi − y ) 2 − βˆ ∑ ( x − x )
2 2
1 i

S yy − SS E = βˆ1 ∑ ( xi − x ) 2
2

βˆ12
Therefore, F0 = = t02
σˆ / S xx
2

Because the square of a t random variable with n − 2 degrees of freedom is an F random variable with
1 and n − 2 degrees of freedom, the usual t-test that compares | t0 | to tα / 2, n − 2 is equivalent to

comparing f 0 = t02 to fα ,1, n − 2 = tα2/ 2,n−2 .

11-44
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

0.9(23)
a) f 0 = = 207 . Reject H 0 : β1 = 0 .
1 − 0.9

23R 2
b) Because f 0.01,1,23 = 7.88 , H0 is rejected if > 7.88 .
1 − R2
That is, H0 is rejected if
23R 2 > 7.88(1 − R 2 )
27.28 R 2 > 7.88
R 2 > 0.289

11-45
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Section 11-8

11-66 a) H 0 : ρ = 0

H1 : ρ ≠ 0 α = 0.05
0.8 20 − 2
t0 = 1− 0.64
= 5.657
t0.025,18 = 2.101
| t0 |> t0.025,18

Reject H0. P-value = (<0.0005)(2) = <0.001

b) H 0 : ρ = 0.5

H1 : ρ ≠ 0.5 α = 0.05

z 0 = (arctanh (0.8) − arctanh ( 0.5) )(17 ) 1 / 2 = 2.265

z .025 = 1.96
| z 0 |> z α / 2
Reject H0. P-value = (0.012)(2) = 0.024.

c) tanh(arctanh 0.8 − z.025

17
) ≤ ρ ≤ tanh(arctanh 0.8 + z.025
17
)

where z.025 = 1.96 . 0 . 5534 ≤ ρ ≤ 0 . 9177 .

Because ρ = 0 and ρ = 0.5 are not in the interval, so reject H0.

11-67 a) H 0 : ρ = 0

H1 : ρ > 0 α = 0.05

0.75 20 − 2
t0 = = 4.81
1− 0.752
t0.05,18 = 1.734
t0 > t0.05,18

Reject H0. P-value < 0.0005

b) H 0 : ρ = 0.5

H1 : ρ > 0.5 α = 0.05

z 0 = (arctanh ( 0.75) − arctanh ( 0.5) )(17 ) 1 / 2 = 1.7467

z .05 = 1.65
z 0 > zα
Reject H0. P-value = 0.04

c) ρ ≥ tanh(arctanh 0.75 − z0.05

17
) where z.05 = 1.64

ρ ≥ 2.26
Because ρ = 0 and ρ = 0.5 are not in the interval, reject the nul hypotheses from parts (a) and (b).

11-68 n = 30 r = 0.83

11-46
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

a) H 0 : ρ = 0

H1 : ρ ≠ 0 α = 0.05
r n−2
t0 = = 0.83 28
= 7.874
1− r 2 1− (0.83) 2

t.025,28 = 2.048
t0 > tα / 2,28

Reject H 0 . P-value = 0.

b) tanh(arctanh 0.83 − z.025

27
) ≤ ρ ≤ tanh(arctanh 0.83 + z.025
27
)

where z.025 = 1.96 . 0.453 ≤ ρ ≤ 1.207 .

a) H 0 : ρ = 0.8

H1 : ρ ≠ 0.8 α = 0.05

z0 = (arctanh 0.83 − arctanh 0.8)(27)1/ 2 = 0.4652

z.025 = 1.96
z0 >/ zα / 2

Do not reject H0. P-value = (0.321)(2) = 0.642.

11-69 n = 50 r = 0.62
a) H 0 : ρ = 0

H1 : ρ ≠ 0 α = 0.01
r n−2
t0 = = 0.62 48
= 5.475
1− r 2 1− (0.62)2

t.005,48 = 2.682
t0 > t0.005,48

Reject H0. P-value ≅ 0

b) tanh(arctanh 0.62 − z.005

47
) ≤ ρ ≤ tanh(arctanh 0.62+ z.005
47
)

where z0.005 = 2.575. 0.3358 ≤ ρ ≤ 0.8007 .

c) Yes.

11-70 a) r = 0.933203
b) H 0 : ρ = 0

H1 : ρ ≠ 0 α = 0.05
r n−2
t0 = = 0.933203 15
1− (0.8709)
= 10.06
1− r 2

t.025,15 = 2.131
t0 > tα / 2,15

Reject H0.

c) yˆ = 0.72538 + 0.498081 x

11-47
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

H 0 : β1 = 0

H1 : β1 ≠ 0 α = 0.05

f 0 = 101 . 16
f .05 ,1 ,15 = 4 . 543
f 0 >> f α ,1 ,15

Reject H0. Conclude that the model is significant at α = 0.05. This test and the one in part b) are
identical.
d) No problems with model assumptions are noted.
Residuals Versus x Residuals Versus the Fitted Values
(response is y) (response is y)

3 3

2 2

1 1

Residual
Residual

0 0

-1 -1

-2 -2

10 20 30 40 50 5 15 25

x Fitted Value

Normal Probability Plot of the Residuals

(response is y)

1
Residual

-1

-2

-2 -1 0 1 2

Normal Score

11-71 a) yˆ = −0.0280411 + 0.990987 x

b) H 0 : β1 = 0

H1 : β1 ≠ 0 α = 0.05

f 0 = 79 . 838
f . 05 ,1 ,18 = 4 . 41
f 0 >> f α ,1 ,18

11-48
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Reject H0.
c) r = 0 . 816 = 0 . 903
d) H 0 : ρ = 0

H1 : ρ ≠ 0 α = 0.05

R n−2 0.90334 18
t0 = = = 8.9345
1− R 2
1 − 0.816
t.025,18 = 2.101
t0 > tα / 2,18

Reject H0.
e) H 0 : ρ = 0.5

H1 : ρ ≠ 0.5 α = 0.05

z0 = 3.879
z.025 = 1.96
z0 > zα / 2

Reject H0.

f) tanh(arctanh 0.90334 − z.025

17
) ≤ ρ ≤ tanh(arctanh 0.90334 + z.025
17
) where z0.025 = 1.96 .

0.7677 ≤ ρ ≤ 0.9615 .

11-72 a) yˆ = 69.1044 + 0.419415 x

b) H 0 : β1 = 0

H1 : β1 ≠ 0 α = 0.05

f 0 = 35 .744
f .05 ,1, 24 = 4 .260
f 0 > f α ,1, 24

Reject H0.
c) r = 0.77349
d) H 0 : ρ = 0

H1 : ρ ≠ 0 α = 0.05

t0 = 0 .77349
1 − 0 . 5983
24
= 5 . 9787
t .025 , 24 = 2 .064
t 0 > t α / 2 , 24

Reject H0.

e) H 0 : ρ = 0.6

H1 : ρ ≠ 0.6 α = 0.05

11-49
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

z0 = (arctanh 0.77349 − arctanh 0.6)(23)1/ 2 = 1.6105

z.025 = 1.96
z0 >/ zα / 2

Do not reject H0.

f) tanh(arctanh 0.77349 − z.025

23
) ≤ ρ ≤ tanh(arctanh 0.77349 + z.025
23
) where z.025 = 1.96 .

0.5513 ≤ ρ ≤ 0.8932 .

11-73 a)

The regression equation is

Current WithoutElect(mA) = 5.50 + 6.73 Supply Voltage

Predictor Coef SE Coef T P

Constant 5.503 3.104 1.77 0.114
Supply Voltage 6.7342 0.7999 8.42 0.000

S = 4.59061 R-Sq = 89.9% R-Sq(adj) = 88.6%

Analysis of Variance

Source DF SS MS F P
Regression 1 1493.7 1493.7 70.88 0.000
Residual Error 8 168.6 21.1
Total 9 1662.3

yˆ = 5.50 + 6.73x

Yes, because the P-value ≈ 0, the regression model is significant at α = 0.05.

b) r = 0.899 = 0.948
c) H 0 : ρ = 0

11-50
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

H1 : ρ ≠ 0
r n−2 0.948 10 − 2
t0 = = = 8.425
1− r2
1 − 0.9482
t0.025,8 = 2.306
t0 = 8.425 > t0.025,8 = 2.306

Reject H0.
⎛ z ⎞ ⎛ z ⎞
d) tanh ⎜ arctan h r − α / 2 ⎟ ≤ ρ ≤ tanh ⎜ arctan h r + α / 2 ⎟
⎝ n−3 ⎠ ⎝ n−3 ⎠

⎛ 1.96 ⎞ ⎛ 19.6 ⎞
tanh ⎜ arctan h 0.948 − ⎟ ≤ ρ ≤ tanh ⎜ arctan h 0.948 + ⎟
⎝ 10 − 3 ⎠ ⎝ 10 − 3 ⎠
0.7898 ≤ ρ ≤ 0.9879

11-74 a)

The regression equation is

Y2001 = - 0.014 + 1.01 Y2000

Predictor Coef SE Coef T P

Constant -0.0144 0.3315 -0.04 0.966
Y2000 1.01127 0.02321 43.56 0.000

S = 0.110372 R-Sq = 99.5% R-Sq(adj) = 99.4%

Analysis of Variance

Source DF SS MS F P
Regression 1 23.117 23.117 1897.63 0.000
Residual Error 10 0.122 0.012
Total 11 23.239

yˆ = −0.014 + 1.011x

Yes, because the P-value ≈ 0, the regression model is significant at α = 0.05.

11-51
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

b) r = 0.995 = 0.9975
c) H 0 : ρ = 0.9

H1 : ρ ≠ 0.9
z0 = (arctan h R − arctan h ρ 0 ) ( n − 3)
1/ 2

z0 = (arctan h 0.9975 − arctan h 0.9) (12 − 3)

1/ 2

z0 = 5.6084
zα / 2 = z0.025 = 1.96
| z0 |> z0.025

Reject H0. P-value = (1 − 1)(2) = 0.000.

⎛ z ⎞ ⎛ z ⎞
d) tanh ⎜ arctan h r − α / 2 ⎟ ≤ ρ ≤ tanh ⎜ arctan h r + α / 2 ⎟
⎝ n−3 ⎠ ⎝ n−3 ⎠

⎛ 1.96 ⎞ ⎛ 19.6 ⎞
tanh ⎜ arctan h 0.9975 − ⎟ ≤ ρ ≤ tanh ⎜ arctan h 0.9975 + ⎟
⎝ 12 − 3 ⎠ ⎝ 12 − 3 ⎠
0.9908 ≤ ρ ≤ 0.9993

11-75 Refer to the Minitab output in Exercise 11-3.

a) r = 0.672 = 0.820
b) H 0 : ρ = 0

H1 : ρ ≠ 0

r n−2 0.82 32 − 2
t0 = = = 7.847
1− r 2
1 − 0.822

t.025,30 = 2.042
t0 > t.025,30

Reject H0. The P-value < 0.0005

⎛ 1.96 ⎞ ⎛ 19.6 ⎞
c) tanh ⎜ arctan h (0.082)− ⎟ ≤ ρ ≤ tanh ⎜ arctan h (0.082)+ ⎟
⎝ 32 − 3 ⎠ ⎝ 32 − 3 ⎠
0.660 ≤ ρ ≤ 0.909

d) H 0 : ρ = 0.7

H1 : ρ ≠ 0.7
z0 = (arctan h R − arctan h ρ0 ) ( n − 3)
1/ 2

z0 = (arctan h 0.82 − arctan h 0.7) ( 32 − 3)

1/ 2

z0 = 1.56
zα / 2 = z0.025 = 1.96
| z0 |< z0.025

Fail to Reject H0. The P-value = 2(0.0594) = 0.1188

11-52
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-76

Here r = 0. The correlation coefficient does not detect the relationship between x and y because the
relationship is not linear. See the graph above.

11-53
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Section 11-9

11-77 a) Yes, ln y = ln β 0 + β1 ln x + ln ε

b) No
c) Yes, ln y = ln β 0 + x ln β1 + ln ε

1 1
d) Yes, = β 0 + β1 + ε
y x

11-78 a) There is curvature in the data.

b) y = −1955.8 + 6.684x
c)
Source DF SS MS F P
Regression 1 491448 491448 35.54 0.000
Residual Error 9 124444 13827
Total 10 615892
d)

There is a curve in the residuals.

11-54
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

e) The data are linear after the transformation to y* = ln y and x* = 1/x.

ln y = 20.6 − 5185(1/x)

Analysis of Variance
Source DF SS MS F P
Regression 1 28.334 28.334 103488.96 0.000
Residual Error 9 0.002 0.000
Total 10 28.336

There is still curvature in the data, but now the plot is convex instead of concave.

11-79 a)

11-55
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

15

10

y
5

0 500 1000 1500 2000 2500 3000 3500

x

b) yˆ = −0.8819 + 0.00385 x

c) H 0 : β1 = 0

H1 : β1 ≠ 0 α = 0.05

f 0 = 122.03
f 0 > f 0.05,1,48

Reject H0. Conclude that regression model is significant at α = 0.05

a) No, it seems the variance is not constant, there is a funnel shape.

Residuals Versus the Fitted Values

(response is y)

0
Residual

-1

-2

-3

-4

-5
0 5 10

Fitted Value

e) yˆ ∗ = 0.5967 + 0.00097 x . Yes, the transformation stabilizes the variance.

11-56
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Section 11-10
1
11-80 a) The fitted logistic regression model is yˆ =
1 + exp[−(−8.84679 − 0.000202 x)]

The Minitab result is shown below

Binary Logistic Regression: Home Ownership Status versus Income

Link Function: Logit
Response Information
Variable Value Count
Home Ownership Status 1 11 (Event)
0 9
Total 20

Logistic Regression Table

Odds 95% CI
Predictor Coef SE Coef Z P Ratio Lower Upper
Constant -8.84679 4.44559 -1.99 0.047
Income 0.0002027 0.0001004 2.02 0.044 1.00 1.00 1.00

Log-Likelihood = -11.163
Test that all slopes are zero: G = 5.200, DF = 1, P-Value = 0.023

b) The P-value for the test of the coefficient of income is 0.044 < α = 0.05. Therefore, income has a
significant effect on home ownership status.
c) The odds ratio is changed by the factor exp(β1) = exp(0.0002027) = 1.000202 for every unit increase
in income. More realistically, if income changes by $1000, the odds ratio is changed by the factor
exp(1000β1) = exp(0.2027) = 1.225.

1
11-81 a) The fitted logistic regression model is yˆ =
1 + exp[ −(5.33968 − 0.000224 x)]

The Minitab result is shown below

Binary Logistic Regression: Number Failing, Sample Size, versus Load (kN/m2)
Link Function: Logit
Response Information
Variable Value Count
Number Failing Failure 337
Success 353
Sample Size Total 690

Logistic Regression Table

Odds 95% CI

11-57
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Predictor Coef SE Coef Z P Ratio Lower Upper

Constant 5.33968 0.545698 9.79 0.000
load (kN/m2) -0.0002246 0.0000228 -9.83 0.000 1.00 1.00 1.00

Log-Likelihood = -421.856
Test that all slopes are zero: G = 112.459, DF = 1, P-Value = 0.000

b) The P-value for the test of the coefficient of load is near zero. Therefore, load has a significant
effect on failing performance.

1
11-82 a) The fitted logistic regression model is yˆ =
1 + exp[ −(−2.12756 + 0.113925 x)]

The Minitab results are shown below

Binary Logistic Regression: Number Redee, Sample size, versus Discount, x

Link Function: Logit

Response Information
Variable Value Count
Number Redeemed Success 2693
Failure 3907
Sample Size Total 6600

Logistic Regression Table

Odds 95% CI
Predictor Coef SE Coef Z P Ratio Lower Upper
Constant -2.12756 0.0746903 -28.49 0.000
Discount, x 0.113925 0.0044196 25.78 0.000 1.12 1.11 1.13

Log-Likelihood = -4091.801
Test that all slopes are zero: G = 741.361, DF = 1, P-Value = 0.000

b) The P-value for the test of the coefficient of discount is near zero. Therefore, discount has a
significant effect on redemption.

11-58
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

c)

d) The P-value of the quadratic term is 0.95 > 0.05, so we fail to reject the null hypothesis of the
quadratic coefficient at the 0.05 level of significance. There is no evidence that the quadratic term is
required in the model. The Minitab results are shown below

Binary Logistic Regression: Number Redee, Sample size, versus Discount, x

Link Function: Logit
Response Information
Variable Value Count
Number Redeemed Event 2693
Non-event 3907
Sample Size Total 6600

Logistic Regression Table

95%
Odds CI
Predictor Coef SE Coef Z P Ratio Lower
Constant -2.34947 0.174523 -13.46 0.000
Discount, x 0.148003 0.0245118 6.04 0.000 1.16 1.11
Discount, x* Discount, x -0.0011084 0.0007827 -1.42 0.157 1.00 1.00

Predictor Upper
Constant
Discount, x 1.22
Discount, x*Discount, x 1.00

Log-Likelihood = -4090.796
Test that all slopes are zero: G = 743.372, DF = 2, P-Value = 0.000

e) The expanded model does not visually provide a better fit to the data than the original model.

Comparison of Data and two Logistic Regression

11-59
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

11-83 a) The Minitab results are shown below

Binary Logistic Regression: y versus Income x1, Age x2

Link Function: Logit
Response Information
Variable Value Count
y 1 10 (Event)
0 10
Total 20

Logistic Regression Table

Odds 95% CI
Predictor Coef SE Coef Z P Ratio Lower Upper
Constant -7.79891 5.05557 -1.54 0.123
Income x1 0.0000833 0.0000678 1.23 0.220 1.00 1.00 1.00
Age x2 1.06263 0.567664 1.87 0.061 2.89 0.95 8.80

Log-Likelihood = -10.423
Test that all slopes are zero: G = 6.880, DF = 2, P-Value = 0.032

b) Because the P-value = 0.032 < α = 0.05 we can conclude that at least one of the coefficients (of
income and age) is not equal to zero at the 0.05 level of significance. The individual z-tests do not
generate P-values less than 0.05, but this might be due to correlation between the independent
variables. The z-test for a coefficient assumes it is the last variable to enter the model. A model
might use either income or age, but after one variable is in the model, the coefficient z-test for the
other variable may not be significant because of their correlation.

c) The odds ratio is changed by the factor exp(β1) = exp(0.0000833) = 1.00008 for every unit increase
in income with age held constant. Similarly, odds ratio is changed by the factor exp(β1) =

11-60
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

exp(1.06263) = 2.894 for every unit increase in age with income held constant. More realistically, if
income changes by $1000, the odds ratio is changed by the factor exp(1000β1) = exp(0.0833) =
1.087 with age held constant.
d) At x1 = 45000 and x2 = 5 from part (a)
1
yˆ = = 0.78
1 + exp[−(−7.79891 + 0.0000833 x1 + 1.06263 x2 )]

e) The Minitab results are shown below

Binary Logistic Regression: y versus Income x1, Age x2

Link Function: Logit
Response Information
Variable Value Count
y 1 10 (Event)
0 10
Total 20

Logistic Regression Table

Odds 95% CI
Predictor Coef SE Coef Z P Ratio Lower Upper
Constant -0.494471 6.64311 -0.07 0.941
Income x1 -0.0001314 0.0001411 -0.93 0.352 1.00 1.00 1.00
Age x2 -2.39447 2.07134 -1.16 0.248 0.09 0.00 5.29
Income x1*Age x2 0.0001017 0.0000626 1.62 0.104 1.00 1.00 1.00

Log-Likelihood = -8.112
Test that all slopes are zero: G = 11.503, DF = 3, P-Value = 0.009

Because the P-value = 0.104 there is no evidence that an interaction term is required in the model.

11-61
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Supplemental Exercises

∑y ∑x
n n n
11-84 a) ∑(y
i =1
i − yˆ i ) = ∑ yi − ∑ yˆ i and
i =1 i =1
i = nβˆ0 + βˆ1 i from the normal equations

Then,
n
(nβˆ0 + βˆ1 ∑ x ) − ∑ yˆ
i =1
i i

n n
= n βˆ0 + βˆ1 ∑
i =1
xi − ∑ (βˆ
i =1
0 + βˆ1 xi )

n n
= n βˆ0 + βˆ1 ∑
i =1
xi − nβˆ0 − βˆ1 ∑x
i =1
i =0

n n n
b) ∑ ( y i − yˆi )xi = ∑ yi xi − ∑ yˆ i xi
i =1 i =1 i =1

n n n
and ∑ yi xi = βˆ0 ∑ xi + βˆ1 ∑ xi 2 from the normal equations. Then,
i =1 i =1 i =1

n n n
βˆ0 ∑ xi + βˆ1 ∑ xi 2 − ∑ (βˆ0 + βˆ1 xi ) xi =
i =1 i =1 i =1
n n n n
βˆ0 ∑ xi + βˆ1 ∑ xi − βˆ0 ∑ xi − βˆ1 ∑ xi 2 = 0
2

i =1 i =1 i =1 i =1

∑ yˆ
1
c) i =y
n i =1

∑ yˆ = ∑ ( βˆ0 +βˆ1 x)

1 n 1
∑ yˆ i = n ∑ ( βˆ 0 + βˆ1 xi )
n i =1
1
= (nβˆ 0 + βˆ1 ∑ xi )
n
1
= (n( y − βˆ1 x ) + βˆ1 ∑ xi )
n
1
= ( ny − nβˆ1 x + βˆ1 ∑ xi )
n
= y − βˆx + βˆ x 1
=y

11-85 a)

11-62
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Plot of y vs x

2.2

1.9

1.6

y
1.3

0.7

1.1 1.3 1.5 1.7 1.9 2.1

Yes, a linear relationship seems plausible.

b) Model fitting results for: y

Independent variable coefficient std. error t-value
sig.level
CONSTANT -0.966824 0.004845 -199.5413
0.0000
x 1.543758 0.003074 502.2588
0.0000
--------------------------------------------------------------------------
R-SQ. (ADJ.) = 1.0000 SE = 0.002792 MAE = 0.002063 DurbWat = 2.843
Previously: 0.0000 0.000000 0.000000 0.000
10 observations fitted, forecast(s) computed for 0 missing val. of dep. var.

yˆ = −0.966824 + 1.54376 x

c) Analysis of Variance for the Full Regression

Source Sum of Squares DF Mean Square F-Ratio P-value
Model 1.96613 1 1.96613 252264. .0000
Error 0.0000623515 8 0.00000779394
----------------------------------------------------------------------------
Total (Corr.) 1.96619 9
R-squared = 0.999968 Stnd. error of est. = 2.79176E-3
R-squared (Adj. for d.f.) = 0.999964 Durbin-Watson statistic = 2.84309

2) H 0 : β1 = 0

3) H1 : β1 ≠ 0

4) α = 0.05
SS R / k
5) The test statistic is f 0 =
SS E /( n − p )

6) Reject H0 if f0 > fα,1,8 where f0.01,1,8 = 11.26

7) Using the results from the ANOVA table

11-63
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

1.96613 /1
f0 = = 252263.9
0.0000623515 / 8
8) Because 252264 > 11.26 reject H0 and conclude that the regression model is significant at α =
0.05.
P-value ≈ 0

d) 99 percent confidence intervals for coefficient estimates

---------------------------------------------------------------------------
Estimate Standard error Lower Limit Upper Limit
CONSTANT -0.96682 0.00485 -0.97800 -0.95565
x 1.54376 0.00307 1.53667 1.55085
---------------------------------------------------------------------------
1.53667 ≤ β1 ≤ 1.55085

e) 2) H 0 : β 0 = 0

3) H1 : β 0 ≠ 0

4) α = 0.01

βˆ0
5) The test statistic is t0 =
se( βˆ0 )

6) Reject H0 if t0 < −tα/2,n-2 where −t0.005,8 = −3.355 or t0 > t0.005,8 = 3.355

7) Using the results from the table above
−0.96682
t0 = = −199.34
0.00485
8) Since −199.34 < −3.355 reject H0 and conclude the intercept is significant at α = 0.05.

11-86 a) yˆ = 93.55 + 15.57 x

b) H 0 : β1 = 0

H1 : β1 ≠ 0

α = 0.05
f 0 = 12.872
f.05,1,14 = 4.60
f 0 > f 0.05,1,14

Reject H0. Conclude that β1 ≠ 0 at α = 0.05.

c) (9.689 ≤ β1 ≤ 21.445)

d) (79.333 ≤ β 0 ≤ 107.767)

e) yˆ = 93.55 + 15.57(2.5) = 132.475

11-64
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

132.475 ± 2.145 136.9 ⎡ 161 + (2.57.017

− 2.325) ⎤
2

⎣ ⎦
132.475 ± 6.49
125.99 ≤ μˆY | x0 = 2.5 ≤ 138.97

11-87 yˆ ∗ = 1.2232 + 0.5075 x where y ∗ = 1/ y . No, the model does not seem reasonable.

The residual plots indicate a possible outlier.

11-88 yˆ = 4.5067 + 2.21517 x , r = 0.992, R2 = 98.43%

The model appears to be an excellent fit. The R2 is large and both regression coefficients are
significant. No, the existence of a strong correlation does not imply a cause and effect relationship.

11-89 yˆ = 0.7916 x

Even though y should be zero when x is zero, because the regressor variable does not usually assume
values near zero, a model with an intercept fits this data better. Without an intercept, the residuals plots
are not satisfactory.

11-90 a)

110

100

90

80
days

70

60

50

40

30
16 17 18
index

b) The regression equation is

yˆ = −193 + 15.296 x

Analysis of Variance
Source DF SS MS F P
Regression 1 1492.6 1492.6 2.64 0.127
Residual Error 14 7926.8 566.2
Total 15 9419.4

11-65
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Do not reject H0. We do not have evidence of a relationship. Therefore, there is not sufficient evidence
to conclude that the seasonal meteorological index (x) is a reliable predictor of the number of days that
the ozone level exceeds 0.20 ppm (y).

c) 95% CI on β1

βˆ1 ± tα / 2 , n − 2 se ( βˆ1 )
15.296 ± t.025 ,12 (9.421)
15.296 ± 2.145(9.421)
− 4.912 ≤ β 1 ≤ 35.504

d) The normality plot of the residuals is satisfactory. However, the plot of residuals versus run order
exhibits a strong downward trend. This could indicate that there is another variable should be
included in the model and it is one that changes with time.

40

30

20

10
Residual

-10

-20

-30

-40

2 4 6 8 10 12 14 16
Observation Order

1
Normal Score

-1

-2
-40 -30 -20 -10 0 10 20 30 40
Residual

11-91 a)

0.7

0.6

0.5
y

0.4

0.3

0.2

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

x

b) yˆ = 0.6714 − 2964 x

11-66
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

c) Analysis of Variance

Source DF SS MS F P
Regression 1 0.03691 0.03691 1.64 0.248
Residual Error 6 0.13498 0.02250
Total 7 0.17189

R2 = 21.47%
Because the P-value > 0.05, reject the null hypothesis and conclude that the model is significant.

d) There appears to be curvature in the data. There is a dip in the middle of the normal probability plot
and the plot of the residuals versus the fitted values shows curvature.

1.5
0.2

1.0

0.1
Normal Score

0.5

Residual
0.0 0.0

-0.5
-0.1

-1.0

-0.2
-1.5
0.4 0.5 0.6
-0.2 -0.1 0.0 0.1 0.2
Fitted Value
Residual

11-92 a)

940
y

930

920

920 930 940

x

b) yˆ = 33.3 + 0.9636 x

c) Predictor Coef SE Coef T P

Constant 66.0 194.2 0.34 0.743
Therm 0.9299 0.2090 4.45 0.002

S = 5.435 R-Sq = 71.2% R-Sq(adj) = 67.6%

Analysis of Variance

Source DF SS MS F P
Regression 1 584.62 584.62 19.79 0.002

11-67
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Residual Error 8 236.28 29.53

Total 9 820.90

Reject the null hypothesis and conclude that the model is significant. Here 77.3% of the variability is
explained by the model.

d) H 0 : β1 = 0

H1 : β1 ≠ 0

α = 0.05

βˆ1 − 1 0.9299 − 1
t0 = = = −0.3354
se( βˆ1 ) 0.2090

ta / 2, n − 2 = t.025,8 = 2.306

Because t0 > −ta / 2, n − 2 , we cannot reject H0 and we conclude that there is not enough evidence to reject

the claim that the devices produce different temperature measurements. Therefore, we assume the
devices produce equivalent measurements.

e) The residual plots to not reveal any major problems.

Normal Probability Plot of the Residuals

(response is IR)

1
Normal Score

-1

-5 0 5

Residual

11-68
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Residuals Versus the Fitted Values

(response is IR)

Residual
0

-5

920 930 940

Fitted Value

11-93 a)

b) yˆ = −0.12 + 1.17 x

c)
Source DF SS MS F P
Regression 1 28.044 28.044 22.75 0.001
Residual Error 8 9.860 1.233
Total 9 37.904

Reject the null hypothesis and conclude that the model is significant.

d) x0 = 4.25 μˆ y| x0 = 4.853

⎛ 1 (4.25 − 4.25) 2 ⎞
4.853 ± 2.306 1.2324 ⎜ + ⎟
⎝ 10 20.625 ⎠
4.853 ± 2.306(0.35106)
4.0435 ≤ μ y| x0 ≤ 5.6625

11-69
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

e) The normal probability plot of the residuals appears linear, but there are some large residuals in the
lower fitted values. There may be some problems with the model.

11-94 a)
The regression equation is
No. Of Atoms (x 10E9) = - 0.300 + 0.0221 power(mW)

Predictor Coef SE Coef T P

Constant -0.29989 0.02279 -13.16 0.000
power(mW) 0.0221217 0.0006580 33.62 0.000

S = 0.0337423 R-Sq = 98.9% R-Sq(adj) = 98.8%

Analysis of Variance

11-70
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Source DF SS MS F P
Regression 1 1.2870 1.2870 1130.43 0.000
Residual Error 13 0.0148 0.0011
Total 14 1.3018

b) Yes, there is a significant regression at α = 0.05 because p-value = 0.000 < α.

c) r = 0.989 = 0.994
d) H 0 : ρ = 0

H1 : ρ ≠ 0
r n−2 0.994 15 − 2
t0 = = = 32.766
1− r2 1 − .9942
t0.025,13 = 2.160
t0 = 32.766 > t0.025,13 = 2.160.

Reject H0. P-value = 0.000.

e) 95% confidence interval for β̂1

βˆ1 ± tα / 2, n − 2 se( βˆ1 )

0.022 ± t0.025,13 (0.00066)
0.022 ± 2.160(0.00066)
0.0206 ≤ βˆ ≤ 0.0234
1

11-95 a)

11-71
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

The relationship between carat and price is not linear. Yes, there is one outlier, observation number 33.

b) The person obtained a very good price—high carat diamond at low price.
c) All the data

The regression equation is

price = - 1696 + 9349 carat

Predictor Coef SE Coef T P

Constant -1696.2 298.3 -5.69 0.000
carat 9349.4 794.1 11.77 0.000

S = 331.921 R-Sq = 78.5% R-Sq(adj) = 77.9%

Analysis of Variance

Source DF SS MS F P
Regression 1 15270545 15270545 138.61 0.000
Residual Error 38 4186512 110171
Total 39 19457057

tα/2,n-2 = t0.005,38 = 2.713

99% confidence interval on β1 .

βˆ1 ± tα / 2, n − 2 se( βˆ1 )

9349 ± t0.005,38 (794.1)
9349 ± 2.713(794.1)
7194.6067 ≤ β1 ≤ 11503.3933

With unusual data omitted

The regression equation is
price_1 = - 1841 + 9809 carat_1

Predictor Coef SE Coef T P

Constant -1841.2 269.9 -6.82 0.000

11-72
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

carat_1 9809.2 722.5 13.58 0.000

S = 296.218 R-Sq = 83.3% R-Sq(adj) = 82.8%

Analysis of Variance
Source DF SS MS F P
Regression 1 16173949 16173949 184.33 0.000
Residual Error 37 3246568 87745
Total 38 19420517

tα/2,n-2 = t0.005,37 = 2.718

99% confidence interval on β1 .

βˆ1 ± tα / 2, n − 2 se( βˆ1 )

9809 ± t0.005,37 (722.5)
9809 ± 2.718(722.5)
7845.25 ≤ β1 ≤ 11772.76

The width for the outlier removed is narrower than for the first case.

11-96
The regression equation is
Population = 3549143 + 651828 Count

Predictor Coef SE Coef T P

Constant 3549143 131986 26.89 0.000
Count 651828 262844 2.48 0.029

S = 183802 R-Sq = 33.9% R-Sq(adj) = 28.4%

Analysis of Variance
Source DF SS MS F P
Regression 1 2.07763E+11 2.07763E+11 6.15 0.029
Residual Error 12 4.05398E+11 33783126799
Total 13 6.13161E+11

yˆ = 3549143 + 651828 x

Yes, the regression is significant at α = 0.01. Care needs to be taken in making cause and effect
statements based on a regression analysis. In this case, it is surely not the case that an increase in the
stork count is causing the population to increase, in fact, the opposite is most likely the case. However,
unless a designed experiment is performed, cause and effect statements should not be made on
regression analysis alone. The existence of a strong correlation does not imply a cause and effect
relationship.

11-73
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

Mind-Expanding Exercises

11-97 The correlation coefficient for the n pairs of data (xi, zi) will not be near unity. It will be near zero. The
data for the pairs (xi, zi) where zi = yi2 will not fall along the line yi = xi which has a slope near unity

and gives a correlation coefficient near unity. These data will fall on a line yi = xi that has a slope

near zero and gives a much smaller correlation coefficient.

S
11-98 a) βˆ1 = xY , βˆ0 = Y − βˆ1 x
S xx

Cov(βˆ0 , βˆ1 ) = Cov(Y , βˆ1 ) − xCov(βˆ1 , βˆ1 )

Cov(Y , SxY ) Cov(∑Yi , ∑Yi ( xi − x )) ∑ ( xi − x )σ
2 − xσ 2
= 0 . Therefore, Cov( β 0 , β1 ) =
ˆ ˆ
Cov(Y , βˆ1 ) = = =
Sxx nSxx nSxx S xx
σ 2
Cov(βˆ1 , βˆ1 ) = V (βˆ1 ) =
Sxx

b) The requested result is shown in part a).

11-99 a) MS E =
∑ (Y − βˆ
i 0 − βˆ1 xi )2
=
∑e 2
i

n−2 n−2

E (ei ) = E (Yi ) − E ( βˆ0 ) − E ( βˆ1 ) xi = 0

( xi − x )2
V (ei ) = σ 2 [1 − ( 1n + S xx
)] Therefore,

E ( MS E ) =
∑ E (e ) = ∑ V (e )
2
i i

n−2 n−2
∑σ
( xi − x ) 2
2
[1 − ( 1n + )]
=
S xx

n−2
σ 2 [n − 1 − 1]
= =σ2
n−2

b) Using the fact that SSR = MSR, we obtain

{
E ( MS R ) = E ( βˆ12 S xx ) = S xx V ( βˆ1 ) + [ E ( βˆ1 )]2 }
⎛σ2 ⎞
= S xx ⎜ + β12 ⎟ = σ 2 + β12 S xx
⎝ xx
S ⎠

11-74
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

S x1Y
11-100 βˆ1 =
S x1 x1

⎡ n ⎤ n
E ⎢ ∑ Yi ( x1i − x1 ) ⎥ ∑ (β 0 + β1 x1i + β 2 x2 i )( x1i − x1 )
E ( βˆ1 ) = ⎣ i =1 ⎦= i =1

S x1 x1 S x1 x1
n
β1 S x x + β 2 ∑ x ( x − x1 )
1 1 2i 1i
β2 Sx x
= i =1
= β1 + 1 2

S x1 x1 S x1 x1

No, β̂1 is no longer unbiased.

σ 2 n
11-101 V ( βˆ1 ) = . To minimize V ( βˆ1 ), S xx should be maximized. Because S xx = ∑ ( xi − x ) 2 , S xx is
S xx i =1

maximized by choosing approximately half of the observations at each end of the range of x.
From a practical perspective, this allocation assumes the linear model between Y and x holds
throughout the range of x and observing Y at only two x values prohibits verifying the linearity
assumption. It is often preferable to obtain some observations at intermediate values of x.

n
11-102 One might minimize a weighted some of squares ∑w (y
i =1
i i − β 0 − β1 xi ) 2 in which a Yi with small variance

(wi large) receives greater weight in the sum of squares.

∂ n n

β0
∑w (y
i =1
i i − β 0 − β1 xi ) 2 = −2∑ wi ( yi − β 0 − β1 xi )
i =1

∂ n n

β1
∑w (y
i =1
i i − β 0 − β1 xi ) 2 = −2∑ wi ( yi − β 0 − β1 xi )xi
i =1

Setting these derivatives to zero yields

βˆ0 ∑ wi + βˆ1 ∑ wi xi = ∑ wi yi

βˆ0 ∑ wi xi + βˆ1 ∑ wi xi 2 = ∑ wi xi yi

and these equations are solved as follows

βˆ 1 =
( ∑ w x y )( ∑ w ) − ∑ w y
i i i i i i

(∑ w )(∑ w x ) − (∑ w x ) 2 2
i i i i i

βˆ0 =
∑ w y − ∑ w x βˆ .
i i i i

∑w ∑w
1
i i

sy
11-103 yˆ = y + r (x − x )
sx

11-75
Applied Statistics and Probability for Engineers, 5th edition February 23, 2010

= y+
S xy ∑(y i − y )2 ( x − x )
S xx S yy ∑ (x − x )
i
2

S xy
= y+ (x − x )
S xx
= y + βˆ1 x − βˆ1 x = βˆ0 + βˆ1 x

∂ n n
11-104 a)
β1
∑(y
i =1
i − β 0 − β1 xi ) 2 = −2∑ ( yi − β 0 − β1 xi )xi
i =1

Upon setting the derivative to zero, we obtain

β 0 ∑ xi + β1 ∑ xi 2 = ∑ xi yi

Therefore,

βˆ1 =
∑ x y − β ∑ x = ∑ x (y − β
i i 0 i i i 0 )
∑x ∑x
2 2
i i

⎛ ∑ xi (Yi − β 0 ) ⎞ ∑ xi 2σ 2 σ2
b) V ( βˆ1 ) = V ⎜ ⎟= =
⎜ ∑ xi ⎟
⎠ ⎡⎣ ∑ xi ⎤⎦ ∑ xi
2 2 2 2
⎝

c) βˆ1 ± tα / 2, n −1 σˆ 2
2
∑ xi

This confidence interval is shorter because ∑ xi ≥ ∑ ( xi − x ) 2 . Also, the t value based on n-1
2

degrees of freedom is slightly smaller than the corresponding t value based on n-2 degrees of
freedom.

11-76