223

Chapter 10
10.1 (a) The sampling distribution of x is approximately normal with mean x = 280 and

60
=
 2.0702 . (b) The mean is 280. One standard deviation from
standard deviation x =
n

840

the mean: 277.9 and 282.1; two standard deviations from the mean: 275.8 and 284.2; and three
standard deviations from the mean: 273.7 and 286.3. Two different sketches are provided below.

Normal density curve

0.20

0.15

0.10

0.05

0.00

273.7

275.8

277.9

280.0
282.1
Sample mean

284.2

286.3

(c) 2 standard deviations; m  4.2 . (d) The confidence intervals drawn will vary, but they
should all have length 2m  8.4 . See the sketch above on the right. (e) About 95% (by the 6895-99.7 rule).
10.2 (a) Incorrect; the probability is either 0 or 1, but we dont know which. (b) Incorrect; the
general form of these confidence intervals is x m , so x will always be in the center of the
confidence interval. (c) Incorrect; the different samples will yield different sample means, and
the distribution of those sample means is used to provide an interval that captures the population
mean. (d) Incorrect; there is nothing magic about the interval from this one sample. Our method
for computing confidence intervals is based on capturing the mean of the population, not a
particular interval from one sample. (e) Correct interpretation.
10.3 No. The student is misinterpreting 95% confidence. This is a statement about the mean
score for all young men, not about individual scores.
10.4 (a) The sampling distribution of x is approximately normal with mean x = and standard

0.4
=
 0.0566 . (b) The mean is . One standard deviation from the mean:
deviation x =
n

50

0.0566 and + 0.0566 ; two standard deviations from the mean: 0.1132 and + 0.1132 ;
and three standard deviations from the mean: 0.1698 and + 0.1698 . Two different sketches
are provided below.

224

Chapter 10

Normal density curve

0.20

0.15

0.10

0.05

0.00

mu-0.1698

mu-0.1132

mu-0.0566

mu

mu+0.0566 mu+0.1132 mu+0.1698

Sample mean

(c) 2 standard deviations; m  0.1132 . (d) About 95% (by the 68-95-99.7 rule). (e) The
confidence intervals drawn will vary, but they should all have length 2m  0.2264 . See the
sketch above on the right.
10.5 (a) 51% 3% or (48%, 54%). (b) 51% is a statistic from one sample. A different sample
may give us a totally different answer. When taking samples from a population, not every
sample of adults will give the same result, leading to sampling variability. A margin of error
allows for this variability. (c) 95% confidence means that this interval was found using a
procedure that produces correct results (i.e., includes the true population proportion) 95% of the
time. (d) Survey errors such as undercoverage or non-response, depending on how the Gallup
Poll was taken, could affect the accuracy. For example, voluntary response will not base the
sample on the population.
10.6 (a) Let xG denote the sample mean for the 10 girls and xB denote the sample mean for the
7 boys. The distribution of xG xB is Normal with mean xG xB = 54.5 54.1 = 0.4 and standard

2.72 2.42
+
 1.25 . Thus,
10
7
0 0.4

P ( xG > xB ) = P ( xG xB > 0 ) = P z >

= P( z > 0.32) = 0.6255 . (b) Generate 10
1.25

observations from a Normal distribution with mean 54.5 inches and standard deviation 2.7
inches. Store the average in a list, say L3. Generate 7 observations from a Normal distribution
with mean 54.1 and standard deviation 2.4 inches. Store the average in another list, say L4.
Repeat the previous steps 200 times. Store the 200 differences L3L4 in list L5. Count how
many of the differences in list L5 are greater than zero. The estimated probability is the count
divided by 200. Clear lists L1 to L5. Generate the means of the 200 samples of 10 girls but the
following commands:
deviation xG xB =

1 C
randNorm(54.5,27,10)L1: mean(L1)L2(C): 1+CC

Continue to press Enter until the counter reaches 200. Now generate the means of the 200
samples of 7 boys:
1 C
randNorm(54.1,2.4,7)L3: mean(L3)L4(C): 1+CC

Continue to press Enter until the counter reaches 200.

Estimating with Confidence

225

Highlight L5 and let L5 = L2 L4. To make it easier to count the number of differences in L5
that are greater than zero, sort L5: SortD(L5)and look to see when the sign changes.

In this simulation 115 out of 200 were greater than zero, or 0.575.
10.7 The figure below (left) shows how much area is in each tail and the value z* you want to
find. Search Table A for 0.0125 (half of the 2.5% that is not included in a 97.5% confidence
interval). This area corresponds to z* = 2.24.

0.4

Normal density curve

Normal density curve

0.4

0.3

0.2

0.975

0.1
0.0125
0.0

0.3

0.1

0.0125
-z*

0
z

z*

0.94

0.2

0.0

0.03
-z*

0.03
0
z

z*

10.8 The figure above (right) shows how much area is in each tail and the value z* you want to
find. Search Table A for 0.03 (half of the 6%% that is not included in a 94% confidence
interval). This area corresponds to z* = 1.88.
10.9 (a) The parameter of interest is = the mean IQ score for all seventh-grade girls in the
school district. The low score (72) is an outlier, but there are no other deviations from
Normality. In addition, the central limit theorem tells us that the sampling distribution of x will
be approximately Normal since n = 31. We are told to treat these 31 girls as an SRS. The 31
measurements in the sample should be independent if there are at least 1031 = 310 seventhgrade girls in this school district. x = 105.84, the 99% confidence interval for , is
15
105.84 2.576
 105.84 6.94 = (98.90, 112.78). With 99% confidence, we estimate the
31

mean IQ score for all seventh-grade girls in the school district to be between 98.90 and 112.78
IQ points. (b) Unless the class was representative of the whole school district, we would not
have been able to generalize our conclusions to the population of interest.

226

Chapter 10

10.10 (a) A pharmaceutical product is a medication or device used to treat patients. This
analysis is important to make sure that the production process is working properly and the
medication contains the correct amount of the active ingredient. (b) The parameter of interest is
= the true concentration of the active ingredient in this specimen. The repeated measurements
are clearly not independent because they are taken on the same specimen. However, we are told
that these repeated measurements follow a Normal distribution and the analysis procedure has no
bias. The sample mean is x = 0.8404 and the 99% confidence interval for , is
0.0068
0.8404 2.576
 0.8404 0.0101 = (0.8303,0.8505) . With 99% confidence, we estimate the
3

true concentration of the active ingredient for this specimen to be between 0.8303 and 0.8505
grams per liter. (c) 99% confident means that if we repeated the entire process of taking three
measurements and computing confidence intervals over and over a large number of times, then
99% of the resulting confidence intervals would contain the true concentration of the active
ingredient.
10.11 (a) We want to estimate = the mean length of time the general managers have spent
with their current hotel chain. The sample of managers is a SRS (stated in question) and the
sample size is large enough (n = 114) to use the central limit theorem to assume normality for the
sampling distribution. The managers length of employment is independent and so the
conditions for a confidence interval for a mean are met. 99% C.I. for =
3.2
11.78 2.576
 11.78 0.77 = (11.01, 12.55). With 99% confidence, we estimate that the
114

mean length of time the general managers have spent with their current hotel chain is between
11.01 and 12.55 years. (b) 46 out of 160 did not reply. This nonresponse could have affected
the results of our confidence interval considerably, especially if those who didnt respond
differed in a systematic way from those who did.
10.12 (a) A scatterplot is shown below (left). There is a very weak, positive association
between the two heights.
70

4
3
2

66
Residual

Height of sister (inches)

68

64

1
0
-1

62

-2
60
-3
64

65

66

67
68
69
70
Height of brother (inches)

71

72

73

64

65

66

67
68
69
70
Height of brother (inches)

71

72

73

(b) Let y = sisters height and x = brothers height. The least squares regression line is
y = 26.74 + 0.5270 x . The slope indicates that the sisters height will increase on average by

Estimating with Confidence

227

0.527 inches for every one inch increase in the brothers height. (c) Tonyas predicted height is
y = 26.74 + 0.5270 70  63.63 inches. (d) No, the residual plot above (right) shows a clear
quadratic pattern and r 2 = 0.311 , so only 31.1% of the variability in the heights of sisters is
explained by the linear regression line using brothers height as the explanatory variable.
10.13 For 80 video monitors the margin of error is m = 1.645

43
 7.9 , which is half of 15.8,
80

the margin of error for n = 20.

10.14 (a) A 98% confidence interval for = the mean scale reading for a 10 gram weight is
0.0002
10.0023 2.33
 10.0023 0.0002 = (10.0021,10.0025) . We are 98% confident that the
5

mean scale reading for a 10 gram weight is between 10.0021 and 10.0025 grams. Notice that 10
is not in the 98% confidence interval, so there is some evidence that the scale might need to be
2
2.33 0.0002
adjusted. (b) To meet the specifications, we need n
 21.7156 or n = 22
0.0001

measurements.
10.15 (a) A 95% confidence interval for = mean score gain on the second SAT Mathematic

50
 22 3.1 = (18.9,25.1) . With 95% confidence we estimate the mean
1000

exam is 22 1.96

gain for all Math SAT second try scores to be between 18.9 and 25.1 points. (b) The 90%
confidence interval is (19.4, 24.6) and the 99% confidence interval is (17.93, 26.07). (c) The
confidence interval widens with increasing confidence level. See the figure below.

Confidence level

99%

95%

90%
17

18

19

20

21
22
23
Mean score gain

24

25

26

10.16 (a) When n = 250, a 95% confidence interval for is

50
22 1.96
 22 6.2 = (15.8,28.2) . (b) (a) When n = 4000, a 95% confidence interval
250
50
for is 22 1.96
 22 1.55 = (20.45,23.55) . (c) The margins of error are 3.099, 6.198,
4000

and 1.550, respectively. The margin of error decreases as the sample size increases (by a factor

228

Chapter 10
2

1.96 50
of 1 n ). (d) To meet the specifications, we need n
= 2401 , so take n = 2401
2

students.

10.17 (a) The computations are correct. (b) Since the numbers are based on a voluntary
response, rather than an SRS, the methods of this section cannot be used; the interval does not
apply to the whole population.
10.18 (a) We dont know if this interval contains the true percent of all voters who favored
Kerry, but it is based on a method that captures the unknown parameter 95% of the time. (b)
Since the margin of error was 2%, the true value of p could be as low as 49%. Thus, the
confidence interval contains some values of p which suggest that Bush will win.
(c) The proportion of voters that favor Kerry is not randomeither a majority favors Kerry,
or they dont. Discussing probabilities about this proportion has little meaning: the probability
the politician asked about is either 1 or 0 (respectively).
10.19 (a) We would want the sample to be a SRS of the population under study and the
observations to be independently sampled. As the sample size is only 25, the population should
be approximately normally distributed. (b) A 95% confidence interval for is 76 12 = (64,
88). We are 95% confident that the population mean is between 64 and 88, within a range of 12
on either side of the sample mean. (c) When using this method for repeated samples, 95% of the
resulting confidence intervals will capture .
10.20 (a) A SRS of all seniors was obtained and the sample size (n = 500) is large enough so
that the distribution of the sample mean will be approximately Normal. The population size is
also clearly greater than 50010 = 5000, so the conditions for using the confidence interval for a
mean are satisfied. A 99% confidence interval for is
100
461 2.576
 461 11.52 = (449.48,472.52) . We are 99% confident that the mean SAT
500

Math score for all California seniors is between 449 and 473 points. (b) In order to estimate the
mean within 5 points, the margin of error needs to be 5 so the sample size must be
2

2.576 100
n
 2654.31 . Take n = 2,655 students.
5

10.21 No. A confidence interval describes a parameter, not an individual measurement.

10.22 The sample sizes and critical values are the same, but the variability in the two
populations is different. The margin of error for the 3rd graders will be smaller because the
variability in the heights of 3rd graders is smaller than the variability in the heights of children in
kindergarten through fifth grade.

Estimating with Confidence

229

10.23 (a) We can be 99% confident that between 63% and 69% of all adults favor such an
0.66 0.34
= 0.03 to find the critical value z*. The critical
amendment. Solve the equation z *
1664
value is z*  2.58 , which means that the confidence level is 99%. (b) The survey excludes
people without telephones, a large percentage of whom would be poor, so this group would be
underrepresented. Also, Alaska and Hawaii are not included in the sample.
10.24 (a) A boxplot (left) and a Normal probability plot (right) are shown below. The median
is almost exactly in the center of the box and the left whisker is only slightly longer than the right
whisker, so the distribution of healing rate appears to be roughly symmetric. The linear trend in
the Normal probability plots indicates that the Normal distribution provides a reasonable model
for the healing rate.

Rate (micrometers per hour)

40
35
30
25
20
15
10
10

15

20
25
30
Rate (micrometers per hour)

35

40

-2

-1

0
Normal scores

8
 25.67 3.10 = (22.57,28.77) . We are
18

(b) A 90% confidence interval for is 25.67 1.645

90% confident that the mean rate of healing is between 22.57 and 28.77 micrometers per hour.
(c) Her interval is wider. To be more confident that an interval contains the true population
parameter, we must increase the length of the interval, which means allow a larger margin of
error. The margin of error for 95% confidence is larger than for 90% confidence.
2

8
1.645 8
1 , so we need n
10.25 We want z
= 173.19 or n = 174 newts.
1
n

10.26 (a) The 90% confidence interval for the mean rate of healing for newts is (22.565,
28.768). The calculator screens are shown below.

230

Chapter 10

(b) For steps 1, 2, and 4 in the Inference Toolbox, see the solution to Exercise 10.11 (a). The
99% confidence interval for the mean number of years for the hotel managers is (11.008,
12.552). The calculator screens are shown below.

10.27 (a) SEM =

s
9.3
= 0.01 , the standard deviation is s  0.0173 .
 1.7898 . (b) Since SEM =
3
27

10.28 (a) df = 11, t * = 1.796 . (b) df = 29, t * = 2.045 . (c) df = 17, t * = 1.333 .
10.29 (a) 0.0228. (b), (c), and (d):
df
P(t > 2) Absolute difference
2
0.0917
0.0689
10
0.0367
0.0139
30
0.0273
0.0045
50
0.0255
0.0027
100 0.0241
0.0013
(e) As the degrees of freedom increase, the area to the right of 2 under the t distributions gets
closer to the area under the standard normal curve to the right of 2.
10.30 (a) The conditions are: SRS, Normality, and Independence. A random sample of n = 8
was taken from a production run. A boxplot (left) and Normal probability plot (right) are shown
below. Even though the left whisker is a little longer than the right whisker, the distribution of
vitamin C level appears to be roughly symmetric. The linear trend in the Normal probability

Estimating with Confidence

231

plots indicates that the Normal distribution is reasonable, even though the sample size is very
small.

Vitamin C (mg/100 g)

30

25

20

15

10
10

15

20
Vitamin C (mg/100 g)

25

30

-1.5

-1.0

-0.5
0.0
Normal scores

0.5

1.0

Since the observations are taken from the same production run, they are not independent.
However, they are representative of the vitamin C content of the CSB produced during this run,
so we will proceed as if the conditions are satisfied. (b) Since n = 8, df = 7 and t * = 2.365 . A
7.1913
 22.5 6.013 = (16.487,28.513) . We are
8

95% confidence interval for is 22.5 2.365

95% confident that the mean vitamin C content of the CSB produced during this run is between
16.487 and 28.513 mg/100 g.
10.31 (a) A histogram (left) and a stemplot (right) are shown below. The distribution is slightly
left-skewed with a center around 19. The gas mileages range from a minimum of 13.6 to a
maximum of 22.6.
Stem-and-leaf of mpg
Leaf Unit = 0.10

Count

14

16

18
20
Gas mileage (mpg)

22

1
4
7
7
8
10
10
7
6
4

13
14
15
16
17
18
19
20
21
22

= 20

6
368
668
2
07
144
9
05
4566

(b) Yes. The sample size is not large enough to use the central limit theorem for Normality.
However, there are no outliers or severe skewness in the sample data that suggest the population
distribution isnt Normal. (c) x = 18.48, s = 3.12 and n = 20, so standard error is 0.6977. Since
n = 20, df = 19 and t* = 2.093, so the margin of error is 1.46. (d) The 95% confidence interval
for is 18.48 1.46 = (17.02, 19.94). With 95% confidence we estimate the mean gas mileage
for this vehicle to be between 17.02 and 19.94 mpg. (e) No, gas mileage depends on driving
habits, and it is unlikely that this one vehicle will be representative of other similar vehicles.

232

Chapter 10

10.32 (a) The histogram (left) and stemplot (right) below show some left-skewness; however,
for such a small sample, the data are not unreasonably skewed. There are no outliers. (b) With
x = 59.59% and s = 6.26% nitrogen, and t* = 2.306 (df = 8), we are 95% confident that the mean
percent of nitrogen in the atmosphere during the late Cretaceous era is between 54.78% and
64.40%.
Stem-and-leaf of nitrogen
Leaf Unit = 1.0

Count

1
2
2
3
3
3
4
(2)
3

52

56

60
Nitrogen (%)

64

4
5
5
5
5
5
6
6
6

= 9

9
1
4

0
33
445

10.33 (a) The histogram (left) and stemplot (right) below show that the distribution is roughly
symmetric with mean x = 3.62 and standard deviation s = 3.055. (b) Using df = 30, t* = 2.042,
and the interval is (2.548, 4.688). Software and the TI calculator gives (2.552, 4.684) using df =
33. With 95% confidence we estimate the mean change in reasoning score after 6 months of
piano lessons for all pre-school children is between 2.55 and 4.68 points. (c) No. We dont know
that students were assigned at random to the groups in this study. Also, some improvement
could come with increased age.
9
8
7

Count

6
5
4
3
2
1
0

-3

3
Scores

Stem-and-leaf of scores
Leaf Unit = 0.10
1
-3 0
3
-2 00
4
-1 0
5
-0 0
6
0
0
7
1
0
10 2
000
15 3
00000
(7) 4
0000000
12 5
00
10 6
000
7
7
00000
2
8
2
9
00

= 34

10.34 (a) Neither the subjects getting the capsules nor the individuals providing them with the
capsules knew which capsules contained caffeine and which were placebos. (b) The differences
(number with caffeine number with placebo) in the number of beats for the 11 subjects are 80,
22, 17, 131, 19, 3, 23, 1, 20, 51, and 3. The histogram (left) and boxplot (right) below
show that the distribution is not symmetric and there are 3 outliers. The mean difference xd =
20.2 is greater than the median difference of 17, and the standard deviation of the differences
sd = 48.75 is much larger than the IQR of 26. (c) No, the t procedure should not be used
because the sample size is small (n = 11) and the differences in beats per minute are clearly not
Normal.

Estimating with Confidence

233

Count

-50

-25

0
25
50
75
Difference (caffeine-placebo)

100

125

-50

50
Difference (caffeine-placebo)

100

150

10.35 (a) Taking d = number of disruptive behaviors on moon days number on other days, we
want to estimate d = the mean difference for dementia patients. We dont know how the
sample was selected. If these patients arent representative of the population of interest, we
wont be able to generalize our results. The sample size is too small (n = 15) for the central limit
theorem to apply, so we examine the sample data. The distribution is roughly symmetric with no
outliers, which gives us no reason to doubt the Normality of the population of differences. We
assume that these 15 difference measurements are independent. xd = 2.43, sd = 1.46, n = 15, and
1.46
= (1.62,
15

t* for df =14 is 2.145. Thus, the 95% confidence interval for d is 2.43 2.145

3.24). On average, the patients have between 1.62 and 3.24 more episodes of aggressive
behavior during moon days than other days. (b) No, this is an observational study; there could
be any number of reasons that there is increased aggressive behavior.
10.36 (a) With data on all U.S. presidents formal inference makes no sense. (b) The 32
students in an AP Statistics class are not a SRS of all students, so the t interval should not be
used to make an inference about the mean amount of time all students spend on the internet. (c)
The stemplot is strongly skewed to the left and the sample size is n = 20, so we cannot trust the t
interval.
10.37 (a) df = 9, t * = 2.262 . (b) df = 19, t * = 2.861 . (c) df =6, t * = 1.440 .
10.38 (a) The histogram (left) and stemplot (right) below show one observation is somewhat
smaller than the others, but it would not be classified as an outlier. The plots do not show severe
skewness or any outliers, so the Normal condition appears to be reasonable based on this small
sample.

234

Chapter 10

2.0

Stem-and-leaf of levels
Leaf Unit = 0.10

= 9

Count

1.5

1
1
1
2
4
(1)
4
4
2

1.0

0.5

0.0

3
4
5
6
Polyphenol level (% change)

0
1
2
3
4
5
6
7
8

5
09
5
04
14

(b) The population of interest is all adults, but since only healthy men were used for this study,
the results can only be generalized to the population of healthy men. In fact, these healthy men
were not randomly selected, so we need to restrict the inference to other healthy men with
characteristics similar to the men in the study. In short, the limitations of inferences based on
observational studies with small samples (of volunteers) are clearly illustrated with this exercise.
The mean of the nine observations is x = 5.5 percent, while s = 2.517 and s 9  0.839 . With
df = 8, the critical value is t* = 1.860 and 90% confidence interval is 5.5 1.8600.839 = 3.939
to 7.061. We are 90% confident that the mean percent change in polyphenol level for healthy
men with characteristics similar to those in this study is between 3.9% and 7.1%. (c) The data
are paired because the polyphenol level for each man is collected before and after the wine
drinking period, so the observations are clearly dependent. However, this is not a matched-pairs
experiment because this group of men only received one treatment.
10.39 Let = the mean HAV angle in the population of all young patients who require HAV
surgery. The t interval may be used (despite the outlier at 50) because n is large (close to 40).
The patients were randomly selected and it is assumed independent. The mean and standard
deviation of the angle of deformity are x = 25.42 and s = 7.475. Using df = 30 and t* = 2.042,
7.475
= (22.944, 27.896). Software and the
38

the 95% confidence interval for is 25.42 2.042

TI calculators use df = 37 and give the interval 22.9642 to 27.8779. With 95% confidence we
estimate the mean HAV angle for all young patients who require HAV surgery to be between 23
and 28.
10.40 (a) Dropping the outlier at 50, we have x = 24.76 and s = 6.34. Using df = 30 and t* =
6.34
= (22.632, 26.888). Software and
37

2.042, the 95% confidence interval for is 24.76 2.042

the TI calculators use df = 36 and give the interval 22.6431 to 26.8704. (b) The interval in part
(a) is narrower than the interval in Exercise 10.39. Removing the outlier decreased the standard
deviation and consequently decreased the margin of error.
10.41 (a) The histogram (left) and boxplot (right) below show that the differences are slightly
left-skewed, with no outliers, so the Normal distribution is reasonable and the t interval should
be reliable. (b) The mean of the differences is xd = 1.45 , the standard deviation is sd = 3.203 ,
and the standard error of the mean is SEM = 0.716. Using df = 19 and t* = 1.729, the 90%

Estimating with Confidence

235
3.203
= (0.212, 2.69). We are 90% confident that the mean
20

confidence interval is 1.45 1.729

increase in listening score after attending the summer institute improves by between 0.212 and
2.69 points. (c) No, their listening skills may have improved for a number of other, for instance
by studying every night or by living with families that only spoke Spanish. There was no control
group for comparison, either.
6

Count

2
1

-4

-2

0
2
4
Difference (posttest-pretest)

-5.0

-2.5

0.0
2.5
Difference (posttest-pretest)

5.0

7.5

10.42 (a) The distribution cannot be normal, because all values must be (presumably) integers
between 0 and 4. (b) The sample size (282) should make the t methods appropriate, because
s
, which
the distribution of ratings can have no outliers. (c) The margin of error is t *
282
is either 0.1611 (Table C) or 0.1591 (software):
df
t*
Interval
Table C
100 2.626 2.22 0.1611 = 2.0589 to 2.3811
Software 281 2.5934 2.22 0.1591 = 2.0609 to 2.3791
We are 99% confident that the mean rating for boys with ADHD is between 2.06 and 2.38 for
this item. (d) Generalizing to boys with ADHD in other locations is not recommended. These
boys were clearly not a random sample from the population of all boys with ADHD.
10.43 These intervals are constructed as in the previous exercise, except for the choice of t*.
Interval
df
t*
90% confidence Table C 100 1.660 2.22 0.1018 = 2.1182 to 2.3218
Software 281 1.6503 2.22 0.1012 = 2.1188 to 2.3212
95% confidence Table C 100 1.984 2.22 0.1217 = 2.0983 to 2.3417
Software 281 1.9684 2.22 0.1207 = 2.0993 to 2.3407
As the confidence level increases, the width of the interval increases.

236

Chapter 10

10.44 (a) The mean difference for these 14 newts is xd = 5.71 , the standard deviation is
10.56
sd = 10.56 , and the standard error of the mean is SEM =
 2.82 . (b) Using df = 13 and t*
14
10.56
or from 10.71 to 0.71
14

= 1.771, the 90% confidence interval is 5.71 1.771

micrometers per hour. If a large number of samples were obtained and the confidence intervals
were computed, then approximately 90% of the intervals would contain the true mean difference
in healing rates. (c) No. A histogram (left) and boxplot (right) are shown below. Since the
sample size (n = 14) is small and the distribution of the difference is skewed to the left with an
outlier, the Normal distribution is not appropriate. The t interval should not be used to make
inferences about experimental-control .
5

Count

-30

-20
-10
0
Difference (experimental-control)

10

-30

-20
-10
0
Difference (experimental-control)

10

10.45 (a) The population is the 175 residents of Tonyas dorm and p is the proportion of all the
residents who like the food. (b) The sample proportion is p =

14
= 0.28 . (c) No, the population is
50

not large enough relative to the sample (N = 175 < 500 = 1050).
10.46 (a) The population is the 2400 students at Glens college, and p is the proportion who
38
= 0.76 . (b) Yeswe have an
believe tuition is too high. (b) The sample proportion is p =
50
SRS, the population is 48 times as large as the sample, and the success count (38) and failure
count (12) are both greater than 10.
10.47 (a) The population is all adult heterosexuals and p is the proportion of all adult
heterosexuals who have had both a blood transfusion and a sexual partner from a high risk of
AIDS group. (b) The sample proportion is p = 0.002. (c) No, there are only 5 or 6 successes
in the sample.
10.48 (a) The standard error of p is SE p = 0.87 0.13 430, 000  0.0005129 . For 99%

confidence, the margin of error is 2.576 SE p  0.001321 . (b) One source of error is indicated
by the wide variation in response rates: We cannot assume that the statements of respondents
represent the opinions of nonrespondents. The effect of the participation fee is harder to predict,
but one possible impact is on the types of institutions that participate in the survey: Even though

Estimating with Confidence

237

the fee is scaled for institution size, larger institutions can more easily absorb it. These other
sources of error are much more significant than sampling error, which is the only error accounted
for in the margin of error from part (a).
10.49 (a) The population is all college undergraduates and the parameter is p = the proportion of
college undergraduates who are abstainers. (b) Example 10.15 states that the sample is an SRS.
The population (all undergraduates) is at least 10 times the sample size of 10,904. The number
of successes np = 2105 and the number of failures n(1 p ) = 8799 are both at least 10, so the
conditions for constructing a confidence interval are satisfied. (c) A 99% confidence interval for
p is 0.193 2.576 0.193 0.807 10,904 = ( 0.183, 0.203) . (d) With 99% confidence, we
estimate between 18.3% and 20.3% of all college undergraduates are classified as abstainers.

1127
 0.6901 or approximate
1633
sample percent = 69% and the margin of error 1.96 0.6901(1 0.6901) 1633  0.022 or 2.2
10.50 The report should include the sample proportion p =

percentage points for 95% confidence. News release: In January 2000, the Gallup Organization
discovered that approximately 69% of adults were satisfied with the way things are going in the
United States. A random sample of 1633 adults participated in the poll and the margin of error is
about 2.2 percentage points. Do you think our readers are still satisfied with the way things are
going in the United States? The results of our local survey will be printed next Wednesday!
10.51 (a) The proportion is p =

15
0.179 0.821
 0.179 and the standard error is SE p =
 0.042 .
84
84

(b) A 90% confidence interval is 0.179 1.645 0.042 = ( 0.110, 0.247 ) . With 90% confidence,
we estimate that between 11.0% and 24.7% of all applicants lie about having a degree.
10.52 (a) The standard error is SE p =

0.54 0.46
 0.0156 , so the 95% confidence interval is
1019

0.54 1.96 0.0156 = ( 0.509, 0.571) . We are 95% confident that the proportion of adults who

would answer Yes is between 50.9% and 57.1%. Notice that 50% is not in the confidence
interval. The margin of error is 1.960.0156 or about 3%, as stated. (b) The sample sizes for
men and women are not provided. (c) The margin of error for women alone would be greater
than 0.03 because the sample size for women alone is smaller than 1019.
10.53 (a) The margins of error are 1.96 p (1 p ) 100 = 0.196 p (1 p ) . See the table below.
(b) With n = 500, the margins of error are 1.96 p (1 p ) 500 . The new margins of error are
less than half their former size (in fact, they have decreased by a factor of

1
5

 0.447 ).

0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
p
(a) m.e. 0.0588 0.0784 0.0898 0.0960 0.0980 0.0960 0.0898 0.0784 0.0588
(b) m.e. 0.0263 0.0351 0.0402 0.0429 0.0438 0.0429 0.0402 0.0351 0.0263

238

Chapter 10

10.54 (a) To meet the specifications, we need 1.96

0.44 0.56
0.03 or
n

1.96
n
0.44 0.56  1051.74 . Take a sample of n = 1052 adults. (b) With the
0.03
2

1.96
conservative guess, we need n
0.5 0.5  1067.11 or 1068 adults. The conservative
0.03
approach requires 16 more adults.

10.55 (a) The 95% confidence interval for p is 0.64 1.96 0.64 0.36 1028  ( 0.61, 0.67 ) .
With 95% confidence we estimate that between 61% and 67% of all teens aged 13 to 17 have
TVs in their rooms. (b) Not all samples will be the same, so there is some variability from
sample to sample. The margin of error accounts for this variability. (c) Teens are hard to reach
and often unwilling to participate in surveys, so nonresponse bias is a major practical difficulty
for this type of poll. Teens can also be sensitive so response bias associated with the wording of
the question or the verbal emphasis by the interviewer may be a problem.
10.56 Our guess is p* = 0.7 , so we need 1.645

0.7 0.3
0.04 or
n

1.645
n
0.7 0.3  355.17 . Take an SRS of n = 356 students. (b) With p = 0.5 and n =
0.04
0.5 0.5
 0.0436 .
356, the margin of error is 1.645
356

171
 0.1943 and the 95% confidence interval is
880
0.1943 1.96 0.1943 0.8057 880  ( 0.1682, 0.2205 ) . We are 95% confident that between
16.82% and 22.05% of all drivers would say that they had run at least one red light. (b) More
than 171 respondents have run red lights. We would not expect very many people to claim they
have run red lights when they have not, but some people will deny running red lights when they
have.
10.57 (a) The sample proportion is p =

10.58 (a) Our guess is p* = 0.2 , so we need 2.576

0.2 0.8
0.015 or
n

2.576
n
0.2 0.8  4718.77 . Take an SRS of n = 4719 customers. (b) With p = 0.1 and
0.015

n = 4719, the margin of error is 2.576

0.1 0.9
 0.0112 .
4719

10.59 (a) The population of interest is all bicyclists aged 15 or older who were fatally injured
and p is the proportion of all fatally injured bicyclists aged 15 or older who tested positive for

Estimating with Confidence

239

alcohol. We do not know that the examined records came from an SRS, so we must be cautious.
Both np = 542 and n(1 p ) = 1169 are at least 10. There are more than 101711 = 17,110 fatally
injured bicyclists in the United States. A 99% confidence interval for p
is 0.317 2.576

0.317 0.683
= ( 0.288,0.346 ) . With 99% confidence, we estimate that between
1711

28.8% and 34.6% of all fatally injured bicyclists aged 15 or older would test positive for alcohol.
(b) No. For example, we do not know what percent of cyclists who were not involved in fatal
accidents had alcohol in their systems. Many other factors, such as not wearing a helmet, need to
be considered.
10.60 Our guess is p* = 0.75 , so we need 1.96

0.75 0.25
0.04 or
n

1.96
n
0.75 0.25  450.19 . Take an SRS of n = 451 Americans with at least one Italian
0.04
grandparent.

390
 0.3275 and a 95% confidence interval for p is
1191
0.3275 1.96 0.3275 0.6725 1191  ( 0.3008, 0.3541) . (b) Only 45 congregations did not
participate in the study, so the nonresponse rate is 45/1236 = 0.0364 or about 3.6%. This
nonresponse rate is quite small, which suggests that the results should be reliable: If we had
information for the few congregations that failed to respond, our conclusions would probably not
change very much. (c) Speakers and listeners probably perceive sermon length differently (just
as, say, students and teachers have different perceptions of the length of a class period).
Listeners tend to exaggerate and report that the sermons lasted longer than they actually did,
while speakers are more likely to report shorter sermons. Since the key informants provided the
information, the estimate of the true population proportion may be too low.
10.61 (a) The sample proportion is p =

10.62 (a) The sample proportion is p =

SE p =

3547
 0.6341 . The standard error is
5594

p (1 p )
 0.00644 , so the margin of error for 90% confidence is 1.645 SE p  0.0106 , and
5594

the interval is 0.6235 to 0.6447. This interval was found using a procedure that includes the true
proportion 90% of the time. (b) Yes, we do not know if those who did respond can reliably
represent those who did not.
10.63 No, the data are not based on an SRS, and therefore inference procedures are not reliable
in this case. A voluntary response sample is typically biased.

107
 0.8425 and a 99% confidence interval for p is
127
0.8425 2.576 0.8425 0.1575 127  ( 0.7592, 0.9258 ) . With 99% confidence, we estimate

10.64 (a) The sample proportion is p =

240

Chapter 10

that between 76% and 93% of all undergraduate students pray at least a few times a year. (b)
No, the fact that these students were all in psychology and communications courses makes it
seem unlikely that they are truly representative of all undergraduates.
CASE CLOSED!
1. Graphical summaries of the call response times are shown below. The histogram (left) and
boxplot (right) clearly show that the distribution of the call response times is strongly skewed to
the right with several outliers. The Normal distribution is certainly not appropriate for these
response times. The mean response time is 31.99 seconds and the median response time is 20
seconds. The large outliers clearly have an impact on the mean, and they will also influence the
standard deviation. The standard deviation is 37.2 seconds and the IQR is 32 seconds.
140
120

Count

100
80
60
40
20
0

60

120
180
240
Call response time (seconds)

300

50

100
150
200
250
Call response time (seconds)

300

350

2. Let denote the mean call response time at the banks customer service center. Using df =
37.2
 (27.73,36.25) .
100 and t* = 1.984, a 95% confidence interval for is 31.99 1.984
300
Software and the TI calculators give the interval from 27.7673 to 36.2194 seconds.
3. Let p denote the proportion of call response times that are at most 30 seconds at the banks
203
 0.6767 and the 95% confidence
customer service center. The sample proportion is p =
300
interval for p is 0.6767 1.96 0.6767 0.3233 300  ( 0.6238, 0.7296 ) .
4. The distribution of response times clearly does not follow a Normal distribution. The major
conditions regarding random sampling and independence are satisfied for the inferences below.
However, it is worth noting that we are relying on the central limit theorem and the robustness of
the t procedures for the inference regarding the mean call response time because the sample size
(n = 300) is large. We are 95% confident that the mean call response time is between
approximately 28 and 36 seconds. The large call response times, which unfortunately occur in
this business, clearly have an impact on the mean. With 95% confidence, we estimate the
proportion of calls answered within 30 seconds to be between 62% and 73%. The intervals
reported above are based on methods that will include the true mean and true proportion for all
calls to your customer service center 95% of the time. (P.S. As you know, there is another way
to describe the center of the call waiting times. This statistic is known as the median and it is
very useful for skewed distributions. If you would like to learn more about inferences based on
the median, we can schedule another meeting.)

Estimating with Confidence

241

10.65 (a) The histogram below shows that the distribution is slightly skewed to the right, but the
Normal distribution is reasonable.
4

Count

223.90

223.95

224.00
224.05
Crankshaft (mm)

224.10

(b) The sample mean is x = 224.002 mm and the standard deviation is s = 0.062 , very close to
the known standard deviation in the population. A 95% confidence interval for is
0.060
224.002 1.96
 ( 223.973, 224.031) . With 95% confidence, we estimate the mean critical
16
dimension for auto engine crankshafts of this type are between 223.973 mm and 224.031 mm.
(c) In repeated samples of this size, 95% of the intervals obtained will contain the true mean.
2
0.06
1.96 0.06
.02 , so we need n
(d) The specification is 1.96
 34.57 or 35 crankshafts.
n
0.02
10.66 (a) If we take a different sample, then we will probably get a different estimate. There is
variability from sample to sample. (b) The sample proportion is p = 0.37 and the 95%
confidence interval for p is 0.37 1.96 0.37 0.63 1000  ( 0.3401, 0.3999 ) . (c) Yes, the
margin of error is 1.96 0.37 0.63 1000  0.0299 or about 3 percentage points. (d) Yes, most
people are not thinking about football during June so the proportion would probably decrease
and more people would say that baseball, tennis, soccer, or racing was their favorite sport to
watch.
10.67 (a) Using df = 26 and t* = 2.779, a 95% confidence interval for is
9.3
114.9 2.779
 (109.93,119.87 ) . With 95% confidence we estimate that the mean seated
27
systolic blood pressure of all healthy white males is between 109.93 and 119.87 mm Hg. (b)
The conditions are SRS, Normality, and Independence. The most important condition is that the
27 members of the placebo group can be viewed as an SRS of the population. The Normality
condition requires that the distribution of seated systolic BP in this population is Normal, or at
least not too nonNormal. Since the sample size is moderate, the procedure should be valid as
long as the data show no outliers and no strong skewness. We must assume that these 27
measurements are independent.
10.68 (a) For each subject, subtract the weight before from the weight after to determine the
weight gain. For example, the weight gain for Subject 1 is 61.7 55.7 = 6 kg. The mean weight

242

Chapter 10

gain for all 16 adults is xd = 4.7313 kg, the standard deviation is sd = 1.7457 kg, and the standard
1.7457
 0.4364 kg. Using df = 15 and t* = 2.131, the 95%
error of the mean is SEM =
16
1.7457
or from 3.8013 to 5.6613 kg. Software and the TI
16

confidence interval is 4.7313 2.131

calculators give the interval (3.8010, 5.6615). (b) Because there are 2.2 kg per pound, multiply
the value in kilograms by 2.2 to obtain pounds. The confidence interval from software and the
calculators becomes 8.3622 to 12.4553 lbs. (c) No, the value 16 is not in our 95% confidence
interval. The data suggest that the excess calories were not converted into weight. The subjects
must have used this energy some other way.

660
= 0.44 and the 95% confidence interval for p is
1500
0.44 1.96 0.44 0.56 1500  ( 0.4149, 0.4651) . With 95% confidence, we estimate that
between 41.5% and 46.5% of all adults would use alternative medicine. (b) The news report
should contain the estimate and the margin of error (0.0251 or 2.51%). A brief, nontechnical
explanation of 95% confidence might also be included. News Release: A nationwide survey
discovered that 44% of adults would use alternative medicine if traditional medicine was not
producing the results they wanted. A random sample of 1500 adults participated in the survey
and the margin of error is about 2.5 percentage points. What percent of our readers do you think
would turn to alternative medicine? The results of our local survey will be printed next Monday!

10.69 (a) The sample proportion is p =

0.8185 0.1815
221
 0.02346 ,
 0.8185 and SE p =
270
270
so the margin of error for a 99% confidence interval is 2.576 0.02346  0.0604 . (b) Using the
0.82 0.18
0.03 or
estimate from part (a) as our guess p* = 0.82, we need 2.576
n
10.70 (a) The sample proportion is p =

2.576
n
0.82 0.18  1088.27 . Take an SRS of n = 1089 doctors. In order to guarantee
0.03
that the margin of error is less than 3 percentage points, the conservative approach with p* = 0.5
2

2.576
should be used. Thus, we would need n
0.5 0.5  1843.27 or 1844 doctors.
0.03

10.71 (a) The histogram below shows that the distribution is skewed to the right and the boxplot
below shows three low and five high outliers.

Estimating with Confidence

243

35
30

Count

25
20
15
10
5
0

10

20

30
Monthly fees ($)

40

50

10

20

30
Monthly fees ($)

40

50

(b) The data are from a random sample, and the sample size is large (n = 50), so the central limit
theorem tells us that the sampling distribution of x is approximately Normal. The population of
commercial Internet service providers is also much larger than 1050 = 500 users. (c) The
sample mean is x = 20.9 , the standard deviation is s = 7.6459 and the standard error of the mean
7.6459
 1.0813 . Using df = 40 and t* = 1.684, the confidence interval for is
is SEM =
50
7.6459
20.9 1.684
= (19.08, 22.72) . Software and the TI calculators give (19.0872, 22.7128) ,
50

using df = 49. With 90% confidence, we estimate the mean cost for users of commercial Internet
service providers in August 2000 to be between $19.08 and $22.72.
10.72 (a) The sample mean is 7.560 = 450 minutes. The margin of error is 20 minutes, so
s
20 40
= 20 minutes. Thus, the standard deviation is s =
1.96
= 64.5363 minutes. (b)
1.96
40
This interpretation is incorrect. The confidence interval provided gives an interval estimate for
the mean lifetime of batteries produced by this company, not individual lifetimes. (c) No, a
confidence interval provides a statement about an unknown population mean, not another sample
mean. (d) We are 95% confident that the mean lifetime of all AA batteries produced by this
company is between 430 and 470 minutes. This interval is based on a method that will capture
the true mean lifetime 95% of the time.

750
0.4202 0.5798
 0.4202 and SE p =
 0.011683 ,
1785
1785
so a 99% confidence interval for p is 0.4202 2.576 0.011683  ( 0.390, 0.450 ) . With 99%
confidence, we estimate that between 39% and 45% of all adults attended church or synagogue
within the last 7 days. (b) Using the conservative guess p* = 0.5 , the specification is
10.73 (a) The sample proportion is p =

0.5 0.5
2.576
2.576
0.01 , so we need n
0.5 0.5 = 16589.44 . Take an SRS of n =
n
0.01
16,590 adults. The use of p* = 0.5 is reasonable because our confidence interval shows that the
actual p is in the range 0.3 to 0.7.

244

Chapter 10

10.74 (a) The differences are spread from 0.018 to 0.020 g, with mean x = 0.0015 g and
standard deviation s = 0.0122 g. A stemplot is shown below, but the sample is too small to make
judgments about skewness or symmetry.
Stem-and-leaf of diff
Leaf Unit = 0.0010
2 -1 85
2 -1
4 -0 65
4 -0
4 0
2
3 0
55
1 1
1 1
1 2
0

= 8

(b) Using df = 7 and t* = 2.365, a 95% confidence interval for is

0.0122
0.0015 2.365
= 0.0015 0.0102 = (0.0117,0.0087) We are 95% confident that the
8

mean difference in TBBMC readings is between 0.0117 and 0.0087 g. (c) The subjects from
this sample may be representative of future subjects, but the test results and confidence interval
are suspect because this is not a random sample.