CHAPTER 11

Section 11-2

11-1. a) yi = β 0 + β 1 xi + ε i
2
S xx = 157.42 − 4314 = 25.348571
S xy = 1697.80 − 43(14572 ) = −59.057143
S xy −59.057143
β1 = = = −2.330
S xx 25.348571
β 0 = y − β1 x = 572 43 ) = 48.013
− ( −2.3298017)( 14
14

b) yˆ = βˆ 0 + βˆ1 x
yˆ = 48.012962 − 2.3298017(4.3) = 37.99
c) yˆ = 48.012962 − 2.3298017(3.7) = 39.39
d) e = y − yˆ = 46.1 − 39.39 = 6.71

11-2. a) yi = β 0 + β1 xi + ε i
2
S xx = 143215.8 − 1478
20
= 33991.6
(1478 )(12.75 )
S xy = 1083.67 − 20
= 141.445
S xy 141.445
βˆ1 = = = 0.00416
S xx 33991.6
βˆ0 = 1220.75 − (0.0041617512)( 1478
20 ) = 0.32999
yˆ = 0.32999 + 0.00416 x
SS E 0.143275
σˆ 2 = MS E = = = 0.00796
n−2 18
0.8
0.7
0.6
0.5
y 0.4
0.3
0.2
0.1
0
-50 0 50 100

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

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

d) βˆ = 0.00416
1

11-1
11-3. a) yˆ = 0.3299892 + 0.0041612( 95 x + 32)
yˆ = 0.3299892 + 0.0074902 x + 0.1331584
yˆ = 0.4631476 + 0.0074902 x
b) βˆ = 0.00749
1

11-4. a)
Regression Analysis - Linear model: Y = a+bX
Dependent variable: Games Independent variable: Yards
--------------------------------------------------------------------------------
Standard T Prob.
Parameter Estimate Error Value Level
Intercept 21.7883 2.69623 8.081 .00000
Slope -7.0251E-3 1.25965E-3 -5.57703 .00001
--------------------------------------------------------------------------------
Analysis of Variance
Source Sum of Squares Df Mean Square F-Ratio Prob. Level
Model 178.09231 1 178.09231 31.1032 .00001
Residual 148.87197 26 5.72585
--------------------------------------------------------------------------------
Total (Corr.) 326.96429 27
Correlation Coefficient = -0.738027 R-squared = 54.47 percent
Stnd. Error of Est. = 2.39287
σˆ 2 = 5.7258
If the calculations were to be done by hand use Equations (11-7) and (11-8).
Regression Plot
y = 21.7883 - 0.0070251 x

S = 2.39287 R-Sq = 54.5 % R-Sq(adj) = 52.7 %

10
y

1500 2000 2500 3000

b) yˆ = 21.7883 − 0.0070251(1800) = 9.143

c) −0.0070251(-100) = 0.70251 games won.
1
d) = 142.35 yds decrease required.
0.0070251
e) yˆ = 21.7883 − 0.0070251(1917) = 8.321
e = y − yˆ
= 10 − 8.321 = 1.679

11-2
11-5. 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
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(7.5) = 38.253

c) yˆ = 13.3202 + 3.32437(5.8980) = 32.9273

yˆ = 32.9273
e = y − yˆ = 30.9 − 32.9273 = −2.0273
d) All the points would lie along the 45% axis 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

11-3
11-6. a)
Regression Analysis - Linear model: Y = a+bX
Dependent variable: Usage Independent variable: Temperature
--------------------------------------------------------------------------------
Standard T Prob.
Parameter Estimate Error Value Level
Intercept -6.3355 1.66765 -3.79906 .00349
Slope 9.20836 0.0337744 272.643 .00000
--------------------------------------------------------------------------------
Analysis of Variance
Source Sum of Squares Df Mean Square F-Ratio Prob. Level
Model 280583.12 1 280583.12 74334.4 .00000
Residual 37.746089 10 3.774609
--------------------------------------------------------------------------------
Total (Corr.) 280620.87 11
Correlation Coefficient = 0.999933 R-squared = 99.99 percent
Stnd. Error of Est. = 1.94284
σˆ 2 = 3.7746
If the calculations were to be done by hand use Equations (11-7) and (11-8).
yˆ = −6.3355 + 9.20836 x
b) yˆ = −6.3355 + 9.20836(55) = 500.124
c) If monthly temperature increases by 1 F, y increases by 9.20836.
d) yˆ = −6.3355 + 9.20836(47) = 426.458
y = 426.458
e = y − yˆ = 424.84 − 426.458 = −1.618

11-7. a)
Predictor Coef StDev T P
Constant 33.535 2.614 12.83 0.000
x -0.03540 0.01663 -2.13 0.047

S = 3.660 R-Sq = 20.1% R-Sq(adj) = 15.7%

Analysis of Variance
Source DF SS MS F P
Regression 1 60.69 60.69 4.53 0.047
Error 18 241.06 13.39
Total 19 301.75

σˆ 2 = 13.392
yˆ = 33.5348 − 0.0353971x
b) yˆ = 33.5348 − 0.0353971(150) = 28.226
c) yˆ = 29.4995
e = y − yˆ = 31.0 − 29.4995 = 1.50048

11-4
11-8. 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 = 46.6041 e = y − yˆ = 1.39592
c) y ˆ = −16.5093 + 0.0693554(950) = 49.38

11-9. a)

9
8
7
6
5
y

4
3
2
1
0

60 70 80 90 100
x

Yes, a linear regression would seem appropriate, but one or two points appear to 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 andyˆ = −10.132 + 0.17429 x

c) yˆ = 4.68265 at x = 85

11-5
11-10. a)

250

200

y
150

100

0 10 20 30 40
x

Yes, a linear regression model appears to be plausable.

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

11-11. 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

11-6
 b) σˆ 2 = 13.81
yˆ = 0.470467 + 20.5673 x

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

d) yˆ = 10.1371 e = 1.6629
11-12. a)

2600
y

2100

1600
0 5 10 15 20 25
x

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

Predictor Coef StDev T P

Constant 2625.39 45.35 57.90 0.000
x -36.962 2.967 -12.46 0.000

S = 99.05 R-Sq = 89.6% R-Sq(adj) = 89.0%

Analysis of Variance
Source DF SS MS F P
Regression 1 1522819 1522819 155.21 0.000
Error 18 176602 9811
Total 19 1699421

b) σˆ 2 = 9811.2
yˆ = 2625.39 − 36.962 x
c) yˆ = 2625.39 − 36.962(20) = 1886.15
d) If there were no error, the values would all lie along the 45 axis. The plot indicates age was reasonable
regressor variable.

2600
2500
2400
2300
2200
FITS1

2100
2000
1900
1800
1700

1600 2100 2600

y

11-7
11-13. βˆ0 + βˆ1 x = ( y − βˆ1 x ) + βˆ1 x = y
11-14. a) The slopes of both regression models will be the same, but the intercept will be shifted.
b) yˆ = 2132.41 − 36.9618 x

βˆ 0 = 2625.39 βˆ 0 ∗ = 2132.41
vs.
βˆ1 = −36.9618 βˆ 1 ∗ = −36.9618

∗
11-15. Let x i = x i − x . Then, the model is Yi ∗ = β 0∗ + β1∗ xi∗ + ε i .
n n
Equations 11-7 and 11-8 can be applied to the new variables using the facts that xi∗ = yi∗ = 0 . Then,
i =1 i =1

βˆ1∗ = βˆ 1 and βˆ 0∗ = 0 .

11-16. The least squares estimate minimizes ( yi − βxi ) 2 . Upon setting the derivative equal to zero, we obtain
2
2 ( yi − βxi ) (− xi ) = 2[ yi xi − β xi ] = 0
yi xi
Therefore, βˆ = 2
.
xi

11-17. 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

11-8
Section 11-5

11-18. a) 1) The parameter of interest is the regressor variable coefficient, β1

2) H 0 :β1 = 0
3) H 1:β1 ≠ 0
4) α = 0.05
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 f0.05,1,12 = 4.75

7) Using results from Exercise 11-1
SS R = βˆ1S 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 > 4.75 reject H 0 and conclude that compressive strength is significant in predicting
intrinsic permeability of concrete at α = 0.05. We can therefore conclude model specifies a useful linear
relationship between these two variables.
P − value ≅ 0.000002
SS E 22.123 σˆ 2 1 . 8436
b) ˆ 2 = MS E =
σ = = 1.8436 and se ( βˆ ) = = = 0 . 2696
n−2 12 1
S xx 25 . 3486

c) se ( βˆ ) = 1 x2 1 3 . 0714 2
0 σˆ 2 + = 1 . 8436 + = 0 . 9043
n S xx 14 25 . 3486
11-19. a) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 :β1 = 0
3) H 1:β1 ≠ 0
4) α = 0.05
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 f0.05,1,18 = 4.414

7) Using the results from Exercise 11-2
SS R = β1S xy = ( 0.0041612)(141.445)
= 0.5886
SS E = S yy − SS R
2
= (8.86 − 1220
.75 ) − 0.5886

= 0.143275
0.5886
f0 = = 73.95
0.143275 / 18
8) Since 73.95 > 4.414, reject H 0 and conclude the model specifies a useful relationship at α = 0.05.

11-9
 P − value ≅ 0.000001
σˆ 2 .00796
b) se( βˆ1 ) = = = 4.8391x10 − 4
S xx 33991.6
1 x 1 73.9 2
se( βˆ 0 ) = σˆ 2 + = .00796 + = 0.04091
n S xx 20 33991.6

11-20. a) Refer to ANOVA table of Exercise 11-4.

1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 :β1 = 0
3) H 1: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,1,26 where f0.01,1,26 = 7.721

7) Using the results of Exercise 10-4
MS R
f0 = = 31.1032
MS E
8) Since 31.1032 > 7.721 reject H 0 and conclude the model is useful at α = 0.01. P − value = 0.000007

b) se ( βˆ ) =
σˆ 2 5 . 7257
1 = = . 001259
S xx 3608611 . 43

ˆ )= 1 x 1 2110 .13 2
se ( β 0
ˆ2
σ + = 5 .7257 + = 2 .6962
n S xx 28 3608611 .43
c) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 : β1 = − 0 .01
3) H 1 : β1 ≠ −0 .01
4) α = 0.01
ˆ + .01
5) The test statistic is t = β 1
0 ˆ )
se( β1

6) Reject H0 if t0 < −tα/2,n-2 where −t0.005,26 = −2.78 or t0 > t0.005,26 = 2.78

7) Using the results from Exercise 10-4
− 0.0070251 + .01
t0 = = 2.3618
0.00125965
8) Since 2.3618 < 2.78 do not reject H 0 and conclude the intercept is not zero at α = 0.01.

11-10
11-21. Refer to ANOVA of Exercise 11-5
a) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 :β1 = 0
3) H 1:β1 ≠ 0
4) α = 0.05, using t-test
β1
5) The test statistic is t 0 =
se(β1)
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 H 0 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) H 1:β1 ≠ 0
4) α = 0.05
MS R SS R / 1
5) The test statistic is f0 = =
MS E SS E / ( n − 2)
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 H 0 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.
c) se ( β
ˆ )= ˆ 2
σ 8 . 7675
1 = = . 39027
S xx 57 . 5631

ˆ )= 1 x 1 6 . 4049 2
se ( β 0
ˆ2
σ + = 8 . 7675 + = 2 . 5717
n S xx 24 57 . 5631

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

2) H 0 : β0 = 0
3) H 1 : β0 ≠ 0
4) α = 0.05, using t-test
ˆ
β 0
5) The test statistic is t0 =
ˆ )
se( β 0
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 H 0 and conclude the intercept is not zero at α = 0.05.

11-11
11-22. Refer to ANOVA for Exercise 10-6
a) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 :β1 = 0
3) H 1:β1 ≠ 0
4) α = 0.01
MS R SS R / 1
5) The test statistic is f0 = =
MS E SS E /(n − 2)
6) Reject H0 if f0 > fα,1,22 where f0.01,1,10 = 10.049
7) Using the results from Exercise 10-6
280583.12 / 1
f0 = = 74334.4
37.746089 / 10
8) Since 74334.4 > 10.049, reject H 0 and conclude the model is useful α = 0.01. P-value < 0.000001
b) se( β̂1 ) = 0.0337744, se( β̂ 0 ) = 1.66765
c) 1) The parameter of interest is the regressor variable coefficient, β1.
2) H 0 :β1 = 10
3) H 1:β1 ≠ 10
4) α = 0.01
βˆ1 − β 1, 0
5) The test statistic is t0 =
se( βˆ1 )
6) Reject H0 if t0 < −tα/2,n-2 where −t0.005,10 = −3.17 or t0 > t0.005,10 = 3.17
7) Using the results from Exercise 10-6
9.21 − 10
t0 = = −23.37
0.0338
8) Since −23.37 < −3.17 reject H 0 and conclude the slope is not 10 at α = 0.01. P-value = 0.
d) H0: β0 = 0 H1: β 0 ≠ 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.

11-23. Refer to ANOVA table of Exercise 11-7

a) H 0 : β1 = 0
H 1 : β1 ≠ 0 α = 0.01
f 0 = 4.53158
f 0.01,1,18 = 8.285
f 0 >/ f α ,1,18
Therefore, do not reject H0. P-value = 0.04734. Insufficient evidence to conclude that the model is a useful
relationship.
b) ˆ ) = 0.0166281
se( β 1
ˆ
se( β ) = 2.61396
0

11-12
 c) H 0 : β1 = −0.05
H 1 : β1 < −0.05
α = 0.01
− 0.0354 − ( −0.05 )
t0 = = 0.87803
0.0166281
t .01,18 = 2.552
t 0 </ −t α ,18
Therefore, do not rejectH0. P-value = 0.804251. Insufficient evidence to conclude that β1 is ≥ -0.05.
d) H0 : β 0 = 0 H 1 : β0 ≠ 0 α = 0.01
t 0 = 12.8291
t .005,18 = 2.878
t 0 > tα / 2,18
Therefore, reject H0. P-value ≅ 0

11-24. Refer to ANOVA of Exercise 11-8

a) H0 : β1 = 0
H1 : β1 ≠ 0
α = 0.05
f 0 = 44.0279
f .05 ,1,11 = 4.84
f 0 > f α ,1,11
Therefore, rejectH0. P-value = 0.00004.
b) ˆ ) = 0.0104524
se( β 1
ˆ
se( β ) = 9.84346
0

c) H 0 : β0 = 0
H1 : β0 ≠ 0
α = 0.05
t0 = −1.67718
t.025 ,11 = 2.201
| t0 |</ −tα / 2 ,11
Therefore, do not rejectH0. P-value = 0.12166.

11-13
11-25. Refer to ANOVA of Exercise 11-9
a) H 0 : β1 = 0
H 1 : β1 ≠ 0
α = 0.05
f 0 = 53.50
f .05,1,18 = 4.414
f 0 > f α ,1,18
Therefore, reject H0. P-value = 0.000009.
b) ˆ ) = 0.0256613
se( β 1
ˆ
se( β ) = 2.13526
0

c) H 0 : β0 = 0
H1 : β0 ≠ 0
α = 0.05
t 0 = - 5.079
t.025,18 = 2.101
| t 0 |> tα / 2,18
Therefore, reject H0. P-value = 0.000078.

11-26. Refer to ANOVA of Exercise 11-11

a) H 0 : β1 = 0
H 1 : β1 ≠ 0
α = 0.01
f 0 = 92.224
f .01,1,16 = 8.531
f 0 > f α ,1,16
Therefore, reject H 0 .
b) P-value < 0.00001
c) ˆ ) = 2.14169
se( β 1
ˆ
se( β ) = 1.93591
0

d) H 0 : β0 = 0
H1 : β0 ≠ 0
α = 0.01
t0 = 0.243
t.005 ,16 = 2.921
t0 >/ tα / 2 ,16
Therefore, do not rejectH0. Conclude, Yes, the intercept should be removed.

11-14
11-27. Refer to ANOVA of Exercise 11-12
a) H 0 : β1 = 0
H 1 : β1 ≠ 0
α = 0.01
f 0 = 155.2
f .01,1,18 = 8.285
f 0 > f α ,1,18
Therefore, reject H0. P-value < 0.00001.
b) ˆ ) = 45.3468
se( β 1
ˆ
se( β ) = 2.96681
0

c) H 0 : β1 = −30
H 1 : β1 ≠ −30
α = 0.01
− 36.9618 − ( −30 )
t0 = = −2.3466
2.96681
t .005 ,18 = 2.878
| t 0 |>/ −t α / 2 ,18
Therefore, do not reject H0. P-value = 0.0153(2) = 0.0306.
d) H 0 : β0 = 0
H1 : β0 ≠ 0
α = 0.01
t0 = 57.8957
t.005 ,18 = 2.878
t 0 > t α / 2,18 , therefore, reject H0. P-value < 0.00001.
e) H0 :β 0 = 2500
H1 : β0 > 2500
α = 0.01
2625.39 − 2500
t0 = = 2.7651
45.3468
t.01,18 = 2.552
t0 > tα ,18 , therefore reject H 0 . P-value = 0.0064.

11-15
 b ˆ
11-28. t0 =
β1 After the transformation βˆ1∗ = β 1 , S xx∗ = a 2 S xx , x ∗ = ax , βˆ0∗ = bβˆ0 , and
σ 2 / S xx a

σˆ ∗ = bσˆ . Therefore, t0∗ = bβˆ1 / a

= t0 .
(bσˆ ) 2 / a 2 S xx

βˆ
11-29. a) has a t distribution with n-1 degree of freedom.
σˆ 2

xi2
b) From Exercise 11-17, βˆ = 21.031461,σˆ = 3.611768, and xi2 = 14.7073 .
The t-statistic in part a. is 22.3314 and H 0 : β 0 = 0 is rejected at usual α values.

| −0.01 − (−0.005) |
11-30. d= = 0.76 , S xx = 3608611.96 .
27
2.4 3608611.96
Assume α = 0.05, from Chart VI and interpolating between the curves for n = 20 and n = 30, β ≅ 0.05 .

Sections 11-6 and 11-7

11-31. tα/2,n-2 = t0.025,12 = 2.179

a) 95% confidence interval on β1 .
βˆ ± t
1 se ( βˆ )
α / 2,n− 2 1

− 2 .3298 ± t .025 ,12 ( 0 .2696 )

− 2 .3298 ± 2 .179 ( 0 .2696 )
− 2 .9173 . ≤ β 1 ≤ −1 .7423 .
b) 95% confidence interval on β0 .
βˆ ± t 0 se ( βˆ ).025 ,12 0

48 . 0130 ± 2 . 179 ( 0 . 5959 )

46 . 7145 ≤ β 0 ≤ 49 .3115 .
c) 95% confidence interval on µ when x 0 = 2.5 .
µˆ Y | x 0 = 48 .0130 − 2 .3298 ( 2 .5) = 42 .1885
( x0 − x ) 2
µˆ Y | x ± t .025 ,12 σˆ 2 ( 1n +
0 S xx
)
( 2 .5 − 3 .0714 ) 2
42 .1885 ± ( 2 .179 ) 1 .844 ( 141 + 25 . 3486
)
42 .1885 ± 2.179 ( 0 .3943 )
41 .3293 ≤ µˆ Y | x 0 ≤ 43 .0477

d) 95% on prediction interval when x 0 = 2.5 .

( x0 − x ) 2
yˆ 0 ± t .025,12 σˆ 2 (1 + 1n + S xx )
( 2.5 − 3.0714 ) 2
42.1885 ± 2.179 1.844 (1 + 141 + 25.348571
)
42.1885 ± 2.179 (1.4056 )
39.1257 ≤ y 0 ≤ 45.2513
It is wider because it depends on both the error associated with the fitted model as well as that with the
future observation.

11-16
11-32. tα/2,n-2 = t0.005,18 = 2.878
a) βˆ1 ± (t 0.005,18 )se( βˆ1 ) .
0.0041612 ± (2.878)(0.000484)
0.0027682 ≤ β1 ≤ 0.0055542
( )
b) βˆ 0 ± t 0.005,18 se( βˆ 0 ) .
0.3299892 ± (2.878)(0.04095)
0.212135 ≤ β 0 ≤ 0.447843
c) 99% confidence interval on µ when x 0 = 85 F .
µˆ Y | x 0 = 0 .683689
( x0 − x ) 2
µˆ Y | x ± t.005 ,18 σˆ 2 ( 1n +
0 S xx
)
( 85 − 73 . 9 ) 2
0 .683689 ± ( 2 .878 ) 0 .00796 ( 201 + 33991 .6
)
0 .683689 ± 0 .0594607
0 .6242283 ≤ µˆ Y | x 0 ≤ 0 .7431497

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

yˆ 0 = 0.7044949
2
(x −x)
yˆ 0 ± t.005,18 σˆ 2 (1 + 1n + 0S xx )
2
− 73.9 )
0.7044949 ± 2.878 0.00796(1 + 201 + (9033991 .6
)
0.7044949 ± 0.2640665
0.4404284 ≤ y0 ≤ 0.9685614

Note for Problems 11-33 through 11-35: These computer printouts were obtained from Statgraphics. For Minitab users, the
standard errors are obtained from the Regression subroutine.

11-33. 95 percent confidence intervals for coefficient estimates

--------------------------------------------------------------------------------
Estimate Standard error Lower Limit Upper Limit
CONSTANT 21.7883 2.69623 16.2448 27.3318
Yards -0.00703 0.00126 -0.00961 -0.00444
--------------------------------------------------------------------------------

a) −0.00961 ≤ β1 ≤ −0.00444.
b) 16.2448 ≤ β0 ≤ 27.3318.
(1800 − 2110.14 ) 2
c) 9.143 ± (2.056) 5.72585( 281 + 3608325 .5
)
9.143 ± 1.2287
7.9143 ≤ µˆ Y | x0 ≤ 10.3717
(1800 − 2110.14 ) 2
d) 9.143 ± (2.056) 5.72585(1 + 281 + 3608325.5
)
9.143 ± 5.0709
4.0721 ≤ y0 ≤ 14.2139

11-17
11-34. 95 percent confidence intervals for coefficient estimates
--------------------------------------------------------------------------------
Estimate Standard error Lower Limit Upper Limit
CONSTANT 13.3202 2.57172 7.98547 18.6549
Taxes 3.32437 0.39028 2.51479 4.13395
--------------------------------------------------------------------------------

a) 2.51479 ≤ β1 ≤ 4.13395.
b) 7.98547 ≤ β0 ≤ 18.6549.
( 7.5 − 6.40492 ) 2
c) 38.253 ± (2.074) 8.76775( 241 + 57.563139
)
38.253 ± 1.5353
36.7177 ≤ µˆ Y | x0 ≤ 39.7883
1 ( 7.5− 6.40492 ) 2
d) 38.253 ± (2.074) 8.76775(1 + 24
+ 57.563139
)
38.253 ± 6.3302
31.9228 ≤ y0 ≤ 44.5832

11-35. 99 percent confidence intervals for coefficient estimates

--------------------------------------------------------------------------------
Estimate Standard error Lower Limit Upper Limit
CONSTANT -6.33550 1.66765 -11.6219 -1.05011
Temperature 9.20836 0.03377 9.10130 9.93154
--------------------------------------------------------------------------------

a) 9.10130 ≤ β1 ≤ 9.31543
b) −11.6219 ≤ β0 ≤ −1.04911
1 + (55−46.5)2
c) 500124
. ± (2.228) 3.774609( 12 3308.9994
)
500.124 ± 1.4037586
498.72024 ≤ µ Y|x 0 ≤ 50152776
.

1 + (55− 46.5)2
d) 500.124 ± (2.228) 3.774609(1 + 12 3308.9994
)
500.124 ± 4.5505644
495.57344 ≤ y0 ≤ 504.67456
It is wider because the prediction interval includes error for both the fitted model and from that associated
with the future observation.

11-36. a) − 0 . 07034 ≤ β 1 ≤ − 0 . 00045

b) 28 . 0417 ≤ β 0 ≤ 39 . 027
c) 28 . 225 ± ( 2 . 101 ) 13 . 39232 ( 201 + ( 150 − 149 . 3 ) 2
48436 . 256
)
28 .225 ± 1 .7194236
26 .5406 ≤ µ y | x 0 ≤ 29 .9794

11-18
 ( 150 − 149 . 3 ) 2
d) 28 . 225 ± ( 2 . 101 ) 13 . 39232 (1 + 1
20
+ 48436 . 256
)
28 . 225 ± 7 . 87863
20 . 3814 ≤ y 0 ≤ 36 . 1386

11-37. a) 0 . 03689 ≤ β 1 ≤ 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 + 1
13
+ 67045 . 97
)
46 . 6041 ± 8 . 779266
37 . 8298 ≤ y 0 ≤ 55 . 3784

11-38. a) 0 . 11756 ≤ β 1 ≤ 0 . 22541

b) − 14 . 3002 ≤ β 0 ≤ − 5 . 32598
( 85 − 82 .3 ) 2
c) 4.76301 ± ( 2.101) 1.982231( 201 + 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 + 1
20
+ 3010 . 2111
)
4 .76301 ± 3 .0345765
1 .7284 ≤ y 0 ≤ 7 .7976

11-39. 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
)
128 .814 ± 16 .980124
111 .8339 ≤ µ y | x 0 ≤ 145 .7941

11-40. a) 14 . 3107 ≤ β 1 ≤ 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 | x 0 ≤ 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

11-41. a) − 43 .1964 ≤ β 1 ≤ − 30 .7272

b) 2530 .09 ≤ β 0 ≤ 2720 . 68
( 20 −13 .3375 ) 2
c) 1886 .154 ± ( 2 . 101 ) 9811 . 21( 201 + 1114 .6618 )
1886 . 154 ± 62 . 370688
1823 . 7833 ≤ µ y | x 0 ≤ 1948 . 5247

11-19
 d) 1886 . 154 ± ( 2 . 101 ) 9811 . 21 (1 + ( 20 −13 . 3375 ) 2
1
20 + 1114 . 6618 )
1886 . 154 ± 217 . 25275
1668 . 9013 ≤ y 0 ≤ 2103 . 4067

Section 11-7

11-42. Use the results of Exercise 11-4 to answer the following questions.
2
a) R = 0.544684 ; The proportion of variability explained by the model.
2 148.87197 / 26
RAdj = 1− = 1 − 0.473 = 0.527
326.96429 / 27
b) Yes, normality seems to be satisfied since the data appear to fall along the straight line.
N o r m a l P r o b a b ility P lo t

99.9

99

95

80
cu m u l.
50
p ercen t
20

0.1

-3 . 9 -1 . 9 0.1 2.1 4.1 6.1

R esid u als

c) Since the residuals plots appear to be random, the plots do not include any serious model inadequacies.
Regression of Games on Yards Residuals vs. Predicted Values

6.1 6.1

4.1 4.1
Residuals
Residuals

2.1 2.1

0.1 0.1

-1.9 -1.9

-3.9 -3.9

1400 1700 2000 2300 2600 2900 0 2 4 6 8 10 12

Yards Predicted Values

11-20
11-43. Use the Results of 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

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% ;
11-44. Use the results of Exercise 11-6 to answer the following questions

11-21
 a) R 2 = 99.986% ; The proportion of variability explained by the model.
b) Yes, normality seems to be satisfied since 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

2
11-45. a) R = 20.1121%
b) These plots indicate presence of outliers, but no real problem with assumptions.
Residuals Versus x
(response is y) ResidualsVersusthe Fitted Values
(responseis y)
10
10
Residual

la
0 u
id
se 0
R

-10 -10
100 200 300 23 24 25 26 27 28 29 30 31
x FittedValue

11-22
 c) The normality assumption appears marginal.
Normal Probability Plot of the Residuals
(response is y)

10

Residual
0

-10
-2 -1 0 1 2

Normal Score

11-46. a)

60

50
y

40

850 950 1050

x

yˆ = 0.677559 + 0.0521753 x
b) H0 : β1 = 0 H 1 : β1 ≠ 0 α = 0.05
f 0 = 7.9384
f.05,1,12 = 4.75
f 0 > fα ,1,12
Reject Ho.
c) σˆ 2 = 25.23842
2
d) σˆ orig = 7.324951
The new estimate is larger because the new point added additional variance not accounted for by the
model.
e) 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.
Normal Probability Plot of the Residuals
(response is y)

10
Residual

-10
-2 -1 0 1 2

Normal Score

11-23
 g) Constant variance assumption appears valid except for the added point.
Residuals Versus the Fitted Values
Residuals Versus x
(response is y)
(response is y)

10
10

Residual
Residual

0 0

-10 -10
850 950 1050 45 50 55

x Fitted Value

11-47. a) R 2 = 71.27%
b) No major departure from normality assumptions.
Normal Probability Plot of the Residuals
(response is y)

1
Residual

-1

-2
-2 -1 0 1 2

Normal Score

c) Assumption of constant variance appears reasonable.

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

x Fitted Value

2
11-48. a) R = 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

11-24
 c) Normality assumption appears reasonable.
Normal Probability Plot of the Residuals
(response is y)

30

20

10

Residual
0

-10

-20

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

Normal Score

11-49. a) R 2 = 85 . 22 %
b) Assumptions appear reasonable, but there is a suggestion that variability increases slightly with y.
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 straight line in the tails of the
normal probability plot.
Normal Probability Plot of the Residuals
(response is y)

5
Residual

-5
-2 -1 0 1 2

Normal Score

11-25
11-50. a) R 2 = 0 . 896081 89% of the variability is explained by the model.

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

Normal Probability Plot of the Residuals

(response is y)

1
Normal Score

-1

-2
-200 -100 0 100

Residual

c) Rn2ew model = 0.9573

Larger, because the model is better able to account for the variability in the data with these two outlying
data points removed.
d) σˆ ol2 d model = 9811 . 21
2
σˆ new model = 4022 . 93
Yes, reduced more than 50%, because the two removed points accounted for a large amount of the error.

( n − 2)(1 − SS
S yy )
E
S yy − SS E S yy − SS E
11-51. Using R 2 = 1 − SS
S yy ,
E
F0 = SS E
= SS E
=
S yy n− 2 σˆ 2
Also,
SS E = ( yi − βˆ 0 − βˆ1 xi ) 2
= ( yi − y − βˆ1 ( xi − x )) 2
( yi − y ) + βˆ1 ( xi − x ) 2 − 2 βˆ1
2
= ( yi − y )(xi − x )
( yi − y ) − βˆ1
2 2 2
= ( xi − x )
S yy − SS E = βˆ1
2 2
( xi − x )
βˆ12
Therefore, F0 =
2
= t 02
σˆ / S xx
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 usually t-test that compares | t0 | to tα / 2, n − 2 is equivalent to comparing f 0 = t02 to

fα ,1, n − 2 = tα / 2, n − 2 .

11-26
11-52. a) f = 0 . 9 ( 23 ) = 207 . Reject H 0 : β 1 = 0 .
0
1 − 0 .9
2
b) Because f . 05 ,1 , 23 = 4 . 28 , H0 is rejected if 23 R > 4 . 28 .
2
1− R
That is, H0 is rejected if
23 R 2 > 4 .28 (1 − R 2 )
27 .28 R 2 > 4 .28
R 2 > 0 .157

11-53. Yes, the larger residuals are easier to identify.

1.10269 -0.75866 -0.14376 0.66992
-2.49758 -2.25949 0.50867 0.46158
0.10242 0.61161 0.21046 -0.94548
0.87051 0.74766 -0.50425 0.97781
0.11467 0.38479 1.13530 -0.82398

11-54. For two random variables X1 and X2,

V ( X 1 + X 2 ) = V ( X 1 ) + V ( X 2 ) + 2 Cov ( X 1 , X 2 )
Then,
V (Yi − Yˆi ) = V (Yi ) + V (Yˆi ) − 2Cov (Yi , Yˆi )
= σ 2 + V ( βˆ 0 + βˆ1 xi ) − 2σ 2 [+
1
n
( xi − x ) 2
S xx ]
=σ 2 +σ 2 [+
1
n
( xi − x )
S xx ]− 2σ
2
[+
2 1
n
( xi − x )
S xx
2
]
[
= σ 2 1 − ( 1n + ( xi − x ) 2
S xx )]
a) Because ei is divided by an estimate of its standard error (when σ2 is estimated by σ2 ), ri has approximate 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.

Section 11-9

11-55. a) yˆ = −0 .0280411 + 0 .990987 x

b) H 0 : β 1 = 0
H 1 : β1 ≠ 0 α = 0.05
f 0 = 79 . 838
f .05 ,1 ,18 = 4 . 41
f 0 >> f α ,1 ,18
Reject H0 .
c) r = 0 . 816 = 0 . 903

11-27
 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
t 0 > t α / 2 ,18
Reject H0 .
e) H 0 : ρ = 0 . 5
H 1 : ρ ≠ 0 .5 α = 0.05
z 0 = 3 . 879
z .025 = 1 . 96
z 0 > zα / 2
Reject H0 .
f) tanh(arcta nh 0.90334 - z.025
17
) ≤ ρ ≤ tanh(arcta nh 0.90334 + z.025
17
) where z.025 = 196
. .
0 . 7677 ≤ ρ ≤ 0 . 9615 .

11-56. a) yˆ = 69 .1044 + 0 . 419415 x

b) H 0 : β 1 = 0
H 1 : 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
0 .77349 24
t0 = 1− 0 .5983
= 5 .9787
t .025 , 24 = 2 .064
t 0 > t α / 2 , 24
Reject H0 .

e) H 0 : ρ = 0 . 6
H 1 : ρ ≠ 0 .6 α = 0.05
z 0 = (arctanh 0 .77349 − arctanh 0 .6 )( 23 ) 1 / 2 = 1 .6105
z .025 = 1 .96
z 0 >/ zα / 2
Do not reject H0 .
f) tanh(arcta nh 0.77349 - z ) ≤ ρ ≤ tanh(arcta nh 0.77349 + ) where z.025 = 196
. .
.025 z .025
23 23

0 . 5513 ≤ ρ ≤ 0 . 8932 .

11-28
11-57. a) r = -0.738027
b) H 0 : ρ = 0
H1 : ρ ≠ 0 α = 0.05
−0.738027 26
t0 = 1− 0.5447
= −5.577
t.025, 26 = 2.056
| t0 |> tα / 2, 26
Reject H0 . P-value = (3.69E-6)(2) = 7.38E-6
c) tanh(arcta nh - 0.738 - z.025
) ≤ ρ ≤ tanh(arcta nh - 0.738 + z.025
)
25 25

where z.025 = 1 .96 . − 0 .871 ≤ ρ ≤ −0 .504 .

d) H 0 : ρ = − 0 .7
H 1 : ρ ≠ − 0 .7 α = 0.05
z 0 = (arctanh − 0.738 − arctanh − 0.7 )( 25 )1 / 2 = −0.394
z.025 = 1.96
| z 0 |< zα / 2
Do not reject H0 . P-value = (0.3468)(2) = 0.6936

1/ 2

11-58 S
R = βˆ1 xx and 1 − R 2 = SS E .
S yy S yy
1/ 2
βˆ1 S xx
n−2 SS E .
Therefore, R n−2 S yy βˆ1 where σˆ 2 =
T0 = = =
1− R2 SS E
1/ 2
σˆ 2 n−2
S yy S xx

11-59 n = 50 r = 0.62
a) H 0 : ρ = 0
H1 : ρ ≠ 0 α = 0.01
r n−2 0 . 62 48
t0 = = = 5 . 475
1− r 2 1 − ( 0 .62 ) 2

t .005 , 48 = 2 . 682
t 0 > t 0 .005 , 48
Reject H0 . P-value ≅ 0
b) tanh(arctanh 0.62 - z ) ≤ ρ ≤ tanh(arctanh 0.62 + z )
.005 .005
47 47

where z.005 = 2.575. 0 .3358 ≤ ρ ≤ 0 . 8007 .

c) Yes.

11-60. n = 10000, r = 0.02

a) H 0 : ρ = 0
H1 : ρ ≠ 0 α = 0.05
r n−2 0.02 10000
t0 = = = 2.0002
1− r 2 1−( 0.02) 2

t.025,9998 = 1.96
t 0 > tα / 2,9998
Reject H0 . P-value = 2(0.02274) = 0.04548

11-29
 b) Since the sample size is so large, the standard error is very small. Therefore, very small differences are
found to be "statistically" significant. However, the practical significance is minimal since r = 0.02 is
essentially zero.

11-61. a) r = 0.933203

b) H0 :ρ = 0
H1 : ρ ≠ 0 α = 0.05
r n−2 0 . 933203 15
t0 = = 1 − ( 0 . 8709 )
= 10 . 06
1− r 2

t.025 ,15 = 2 . 131

t 0 > tα / 2 ,15
Reject H0.

c) yˆ = 0.72538 + 0.498081x
H 0 : β1 = 0
H 1 : 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) H0 : β0 = 0
H1 : β0 ≠ 0 α = 0.05
t 0 = 0 . 468345
t . 025 ,15 = 2 . 131
t 0 >/ t α / 2 , 15

Do not reject H0. We cannot conclude β0 is different from zero.

e) 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

11-30
11-62. n = 25 r = 0.83
a) H 0 : ρ = 0
H 1 : ρ ≠ 0 α = 0.05
r n−2 0 .83 23
t0 = = = 7 . 137
1− r 2 1− ( 0 . 83 ) 2

t.025 , 23 = 2 . 069
t 0 > tα / 2 , 23
Reject H0 . P-value = 0.
z.025 z.025
b) tanh(arctanh 0.83 - ) ≤ ρ ≤ tanh(arctanh 0.83 + )
22 22

. . 0 . 6471 ≤ ρ ≤ 0 . 9226 .
where z.025 = 196
c) H 0 : ρ = 0. 8
H 1 : ρ ≠ 0 .8 α = 0.05
z 0 = (arctanh 0.83 − arctanh 0.8)( 22 )1 / 2 = 0.4199
z.025 = 1.96
z 0 >/ zα / 2
Do not reject H 0 . P-value = (0.3373)(2) = 0.6746.

Supplemental Exercises

n n n
11-63. a) ( y i − yˆ i ) = yi − yˆ i and yi = nβˆ 0 + βˆ1 xi from normal equation
i =1 i =1 i =1
Then,
n
(nβˆ 0 + βˆ1 xi ) − yˆ i
i =1
n n
= nβˆ 0 + βˆ1 xi − ( βˆ0 + βˆ1 xi )
i =1 i =1
n n
= nβˆ 0 + βˆ1 xi − nβˆ 0 − βˆ1 xi = 0
i =1 i =1
n n n
b) ( y i − yˆ i )xi = yi xi − yˆ i xi
i =1 i =1 i =1
n n n
and yi xi = βˆ0 xi + βˆ1 xi 2 from 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 2 − βˆ0 xi − βˆ1 xi 2 = 0
i =1 i =1 i =1 i =1

11-31
 n
1
c) yˆ i = y
n i =1

yˆ = ( βˆ0 +βˆ1 x)

1 n 1
yˆ i = ( βˆ 0 + βˆ1 xi )
n i=1 n
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

11-64. a)
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 straight line relationship seems plausible.

11-32
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.

y = −0.966824 + 154376
. 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) H 1:β1 ≠ 0
4) α = 0.05
SS R / k
5) The test statistic is f0 =
SS E /(n − p )
6) Reject H0 if f0 > fα,1,8 where f0.05,1,8 = 5.32
7) Using the results from the ANOVA table
1.96613 / 1
f0 = = 255263.9
0.0000623515 / 8
8) Since 2552639 > 5.32 reject H0 and conclude that the regression model is significant at α = 0.05.
P-value < 0.000001
d) 95 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
--------------------------------------------------------------------------------

−0.97800 ≤ β 0 ≤ −0.95565

e) 2) H 0 :β 0 = 0
3) H 1:β 0 ≠ 0
4) α = 0.05
β0
5) The test statistic is t 0 =
se(β0 )
6) Reject H0 if t0 < −tα/2,n-2 where −t0.025,8 = −2.306 or t0 > t0.025,8 = 2.306
7) Using the results from the table above
−0.96682
t0 = = −199.34
0.00485
8) Since −199.34 < −2.306 reject H 0 and conclude the intercept is significant at α = 0.05.

11-33
11-65. a) yˆ = 93.34 + 15.64 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) (7.961 ≤ β1 ≤ 23.322)
d) (74.758 ≤ β 0 ≤ 111.923)
e) yˆ = 93.34 + 15.64(2.5) = 132.44

132.44 ± 2.145 136.27 161 + ( 2.57−.2017

.325 )
[ 2
]
132.44 ± 6.47
125.97 ≤ µˆ Y |x0 = 2.5 ≤ 138.91

11-66 a) There is curvature in the data.

800
15 700
Vapor Pressure (mm Hg)

600

500
10 400

300
y

200
5
100

280 3 30 38 0
0 T em p erature (K )
0 500 1000 1500 2000 2500 3000 3500
x

b) y = - 1956.3 + 6.686 x

c) Source DF SS MS F P
Regression 1 491662 491662 35.57 0.000
Residual Error 9 124403 13823
Total 10 616065

11-34
d)

Residuals Versus the Fitted Values

(response is Vapor Pr)

200

100
Residual

-100

-200 -100 0 100 200 300 400 500 600

Fitted Value

There is a curve in the residuals.

e) The data are linear after the transformation.

5
Ln(VP)

1
0 .0 0 2 7 0 .0 0 3 2 0 .0 0 3 7
1/T

lny = 20.6 - 5201 (1/x)

Analysis of Variance
Source DF SS MS F P
Regression 1 28.511 28.511 66715.47 0.000
Residual Error 9 0.004 0.000
Total 10 28.515

11-35
 Residuals Versus the Fitted Values
(response is y*)

0.02

0.01

Residual 0.00

-0.01

-0.02

-0.03

1 2 3 4 5 6 7

Fitted Value

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

11-67. a)

15

10
y

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

11-36
 d) No, it seems the variance is not constant, there is a funnel shape.
Residuals Versus the Fitted Values
(response is y)

Residual
-1

-2

-3

-4

-5
0 5 10

Fitted Value

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

1
11-68. yˆ ∗ = 1.2232 + 0.5075 x where y∗ = . No, model does not seem reasonable. The residual plots indicate a
y
possible outlier.

11-69. yˆ = 0.7916 x
Even though y should be zero when x is zero, because the regressor variable does not normally assume values
near zero, a model with an intercept fits this data better. Without an intercept, the MSE is larger because there are fewer
terms and the residuals plots are not satisfactory.

2
11-70. yˆ = 4.5755 + 2.2047 x , r= 0.992, R = 98.40%
2
The model appears to be an excellent fit. Significance of regressor is strong and R is large. Both regression
coefficients are significant. No, the existence of a strong correlation does not imply a cause and effect
relationship.

11-71 a)

110

100

90

80
days

70

60

50

40

30
16 17 18
index

11-37
 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

Cannot reject Ho; therefore we conclude that the model is not significant. Therefore the seasonal
meteorological index (x) is not 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, one that
changes with time.

2 40

30

1 20
Normal Score

10
Residual

0 0

-10

-1 -20

-30
-2
-40
-40 -30 -20 -10 0 10 20 30 40
Residual 2 4 6 8 10 12 14 16
Observation Order

11-72 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

11-38
 b) yˆ = .6714 − 2964 x

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%

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

11-73 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 z i = y i2 will not fall along the straight line y i = xi which has a slope near unity and gives a

correlation coefficient near unity. These data will fall on a line y i = xi that has a slope near zero and gives a much
smaller correlation coefficient.

11-74 a)

5
y

1
2 3 4 5 6 7
x

b) yˆ = −0.699 + 1.66 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

11-39
 d) x=4.25 µ y| x = 4.257
0

1 (4.25 − 4.75) 2
4.257 ± 2.306 1.2324 +
10 20.625
4.257 ± 2.306(0.3717)

3.399 ≤ µ y|xo ≤ 5.114

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

1
Normal Score

Residual
0
0

-1
-1

-2
2 3 4 5 6 7 8
-2 -1 0 1 2
Fitted Value
Residual

11-75 a)

940
y

930

920

920 930 940

x

b) yˆ = 33.3 + 0.9636 x

11-40
 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
Residual Error 8 236.28 29.53
Total 9 820.90

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

d) H 0 : β1 = 1
H 1 : β1 ≠ 1 α=.05

βˆ1 − 1 0.9299 − 1
t0 = = = −0.3354
se( βˆ1 ) 0.2090
t a / 2,n − 2 = t .025,8 = 2.306

Since t 0 > −t a / 2,n − 2 , we cannot reject Ho 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

11-41
 Residuals Versus the Fitted Values
(response is IR)

5
Residual

-5

920 930 940

Fitted Value

Mind-Expanding Exercises

S xY
11-76. a) β̂ 1 = , βˆ 0 = Y − βˆ 1 x
S xx
Cov(βˆ0 , βˆ1 ) = Cov(Y , βˆ1 ) − xCov(βˆ1, βˆ1 )
Cov(Y , S xY ) Cov( Yi , Yi ( xi − x )) ( xi − x )σ 2
Cov(Y , βˆ1 ) = = = = 0. Therefore,
Sxx nSxx nSxx
2
σ
Cov(βˆ1, βˆ1 ) = V (βˆ1 ) =
S xx
− xσ 2
Cov( βˆ0 , βˆ1 ) =
S xx
b) The requested result is shown in part a.

(Yi − βˆ0 − βˆ1 xi ) 2 ei2

11-77. a) MS E = =
n−2 n−2
ˆ ˆ
E (ei ) = E (Yi ) − E ( β 0 ) − E ( β1 ) xi = 0
( xi − x ) 2
V (ei ) = σ 2 [1 − ( 1n + S xx
)] Therefore,

E (ei2 ) V (ei )
E ( MS E ) = =
n−2 n−2
2
σ [1 − ( n1 + ( x S− x ) )]
2 i

= xx

n−2
2
σ [n − 1 − 1]
= =σ2
n−2

11-42
 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
S xx

S x1Y
11-78. βˆ1 =
S x1 x1
n n
E Yi ( x1i − x1 ) ( β 0 + β1x1i + β 2 x2 i )( x1i − x1 )
E ( βˆ1 ) = i =1
= i =1
S x1 x1 S x1 x1
n
β1S x x + β 2 x2 i ( x1i − x1 )
1 1
i =1
β2Sx x
= = β1 + 1 2

S x1 x1 S x1 x1
No, β1 is no longer unbiased.

2
σ n
11-79. V ( βˆ1 ) = . To minimize V ( βˆ1 ), S xx should be maximized. Because S xx = ( xi − x ) 2 , Sxx 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
11-80. One might minimize a weighted some of squares wi ( yi − β 0 − β1 xi ) 2 in which a Yi with small variance
i =1
( w i large) receives greater weight in the sum of squares.

∂ n n
wi ( yi − β 0 − β1 xi )2 = −2 wi ( yi − β 0 − β1 xi )
β0 i =1 i =1

∂ n n
wi ( yi − β 0 − β1xi )2 = −2 wi ( yi − β 0 − β1 xi )xi
β1 i =1 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
as requested.

11-43
 and
( wi xi y i )( wi ) − wi y i
βˆ 1=
( wi ) ( wi xi2 − ( ) wi xi )
2

wi y i wi xi ˆ
βˆ 0 = − β1 .
wi wi

sy
11-81. yˆ = y + r (x − x)
sx
S xy ( yi − y ) 2 ( x − x )
= y+
S xx S yy ( xi − x ) 2
S xy
= y+ (x − x)
S xx
= y + βˆ1 x − βˆ1 x = βˆ0 + βˆ1 x

∂ n n
11-82. a) ( yi − β 0 − β1 xi ) 2 = −2 ( yi − β 0 − β1 xi )xi
β1 i =1 i =1
Upon setting the derivative to zero, we obtain
2
β0 xi + β1 xi = xi yi
Therefore,
xi yi − β 0 xi xi ( yi − β 0 )
βˆ1 = 2
= 2
xi xi

2
xi (Yi − β 0) xi σ 2 σ2
b) V ( βˆ1 ) = V 2
= 2
= 2
xi [ xi ]2 xi

c) βˆ1 ± tα / 2, n−1 σˆ 2
xi 2

This confidence interval is shorter because xi 2 ≥ ( xi − x) 2 . Also, the t value based on n-1 degrees of freedom is
slightly smaller than the corresponding t value based on n-2 degrees of freedom.

11-44