9 A Estimation Using a Single Sample

Document technical information

Format pdf
Size 4.3 MB
First found Jun 9, 2017

Document content analysis

Category Also themed
Language
English
Type
not defined
Concepts
no text concepts found

Persons

Jacob Lawrence
Jacob Lawrence

wikipedia, lookup

Chen Cheng
Chen Cheng

wikipedia, lookup

Organizations

Places

Transcript

Chapter
9
© George Hall /Corbis
Estimation Using
a Single Sample
A
ffirmative action in university admissions is a controversial topic. Some believe
that affirmative action programs are no longer needed, whereas others argue that using race and ethnicity as factors in university admissions is necessary to achieve diverse
student populations. To assess public opinion on this issue, investigators conducted a
survey of 1013 randomly selected U.S. adults. The results are summarized in the article
“Poll Finds Sharp Split on Affirmative Action” (San Luis Obispo Tribune, March 8,
2003). The investigators wanted to use the survey data to estimate the true proportion
of U.S. adults who believed that programs that give “advantages and preferences to
Blacks, Hispanics, and other minorities in hiring, promotions, and college admissions”
should be continued. The methods introduced in this chapter will be used to produce
the desired estimate. Because the estimate is based only on a sample rather than on a
census of all U.S. adults, it is important that this estimate be constructed in a way that
also conveys information about the anticipated accuracy.
The objective of inferential statistics is to use sample data to decrease our uncertainty about some characteristic of the corresponding population, such as a population
mean m or a population proportion p. One way to accomplish this uses the sample data
to arrive at a single number that represents a plausible value for the characteristic of
interest. Alternatively, an entire range of plausible values for the characteristic can be
reported. These two estimation techniques, point estimation and interval estimation,
are introduced in this chapter.
Improve your understanding and save time! Visit http://www.thomsonedu.com/login where you will find:
■
■
■
Step-by-step instructions for MINITAB, Excel, TI-83, SPSS,
and JMP
Video solutions to selected exercises
Data sets available for selected examples and exercises
■
■
Exam-prep pre-tests that build a Personalized Learning
Plan based on your results so that you know exactly
what to study
Help from a live statistics tutor 24 hours a day
475
476
Chapter 9
■
Estimation Using a Single Sample
........................................................................................................................................
9.1
Point Estimation
The simplest approach to estimating a population characteristic involves using sample
data to compute a single number that can be regarded as a plausible value of the characteristic. For example, sample data might suggest that 1000 hr is a plausible value
for m, the true mean lifetime for lightbulbs of a particular brand. In a different setting,
a sample survey of students at a particular university might lead to the statement that
.41 is a plausible value for p, the true proportion of students who favor a fee for recreational facilities.
D
E F I N I T I O N
A point estimate of a population characteristic is a single number that is
based on sample data and represents a plausible value of the characteristic.
In the examples just given, 1000 is a point estimate of m and .41 is a point estimate of p. The adjective point reflects the fact that the estimate corresponds to a single
point on the number line.
A point estimate is obtained by first selecting an appropriate statistic. The estimate
is then the value of the statistic for the given sample. For example, the computed value
of the sample mean provides a point estimate of a population mean m.
..........................................................................................................................................
Example 9.1
Support for Affirmative Action
One of the purposes of the survey on affirmative action described in the chapter introduction was to estimate the proportion of the U.S. population who believe that affirmative action programs should be continued. The article reported that 537 of the
1013 people surveyed believed that affirmative action programs should be continued.
Let’s use this information to estimate p, where p is the true proportion of all U.S.
adults who favor continuing affirmative action programs. With success identified as a
person who believes that affirmative action programs should continue, p is then just
the population proportion of successes. The statistic
p
number of successes in the sample
n
which is the sample proportion of successes, is an obvious choice for obtaining a
point estimate of p. Based on the reported information, the point estimate of p is
p
537
.530
1013
That is, based on this random sample, we estimate that 53% of the adults in the
United States believe that affirmative action programs should be continued.
■
For purposes of estimating a population proportion p, there is no obvious alternative
to the statistic p. In other situations, such as the one illustrated in Example 9.2, there
may be several statistics that can be used to obtain an estimate.
9.1
■
Point Estimation 477
..........................................................................................................................................
Example 9.2
Internet Use by College Students
●
The article “Online Extracurricular Activity” (USA Today, March 13, 2000) reported the results of a study of college students conducted by a polling organization
called The Student Monitor. One aspect of computer use examined in this study was
the number of hours per week spent on the Internet. Suppose that the following observations represent the number of Internet hours per week reported by 20 college
students (these data are compatible with summary values given in the article):
4.00 5.00 5.00 5.25 5.50 6.25 6.25 6.50 6.50
7.00
7.25 7.75 8.00 8.00 8.00 8.25 8.50 8.50 9.50 10.50
A dotplot of the data is shown here:
4
5
6
7
8
9
10
Internet time
If a point estimate of m, the true mean Internet time per week for college students, is desired, an obvious choice of a statistic for estimating m is the sample mean
x. However, there are other possibilities. We might consider using a trimmed mean
or even the sample median, because the data set exhibits some symmetry. (If the
corresponding population distribution is symmetric, the population mean m and the
population median are equal).
The three statistics and the resulting estimates of m calculated from the data are
gx
141.50
7.075
n
20
7.0 7.25
sample median 7.125
2
112.5
average of middle
7.031
b 10% trimmed mean a
16 observations
16
sample mean x The estimates of the mean Internet time per week for college students differ somewhat from one another. The choice from among them should depend on which statistic tends, on average, to produce an estimate closest to the true value of m. The
following subsection discusses criteria for choosing among competing statistics.
■
■
Choosing a Statistic for Computing an Estimate .........................................
The point of Example 9.2 is that more than one statistic may be reasonable to use to
obtain a point estimate of a specified population characteristic. Loosely speaking, the
statistic used should be one that tends to yield an accurate estimate—that is, an estimate close to the value of the population characteristic. Information about the accuracy of estimation for a particular statistic is provided by the statistic’s sampling distribution. Figure 9.1 displays the sampling distributions of three different statistics. The
Step-by-step technology instructions available online
● Data set available online
478
Chapter 9
■
Estimation Using a Single Sample
value of the population characteristic, which we refer to as the true value, is marked
on the measurement axis.
The distribution in Figure 9.1(a) is that of a statistic unlikely to yield an estimate
close to the true value. The distribution is centered to the right of the true value, making it very likely that an estimate (a value of the statistic for a particular sample) will
be larger than the true value. If this statistic is used to compute an estimate based on
a first sample, then another estimate based on a second sample, and another estimate
based on a third sample, and so on, the long-run average value of these estimates will
exceed the true value.
Sampling
distributions of three different statistics for estimating a
population characteristic.
Figure 9.1
True
value
True
value
(a)
(b)
True
value
(c)
The sampling distribution of Figure 9.1(b) is centered at the true value. Thus, although one estimate may be smaller than the true value and another may be larger,
when this statistic is used many times over with different samples, there will be no
long-run tendency to over- or underestimate the true value. Note that even though the
sampling distribution is correctly centered, it spreads out quite a bit about the true
value. Because of this, some estimates resulting from the use of this statistic will be
far above or far below the true value, even though there is no systematic tendency to
underestimate or overestimate the true value.
In contrast, the mean value of the statistic with the distribution shown in Figure 9.1(c) is equal to the true value of the population characteristic (implying no systematic error in estimation), and the statistic’s standard deviation is relatively small.
Estimates based on this third statistic will almost always be quite close to the true
value—certainly more often than estimates resulting from the statistic with the sampling distribution shown in Figure 9.1(b).
D
E F I N I T I O N
A statistic whose mean value is equal to the value of the population characteristic being estimated is said to be an unbiased statistic. A statistic that is
not unbiased is said to be biased.
As an example of a statistic that is biased, consider using the sample range as an
estimate of the population range. Because the range of a population is defined as the
difference between the largest value in the population and the smallest value, the range
for a sample tends to underestimate the population range. This is true because the
9.1
■
Point Estimation 479
largest value in a sample must be less than or equal to the largest value in the population and the smallest sample value must be greater than or equal to the smallest value
in the population. The sample range equals the population range only if the sample includes both the largest and the smallest values in the population; in all other instances,
the sample range is smaller than the population range. Thus, msample range population
range, implying bias.
Let x1, x2, ..., xn represent the values in a random sample. One of the general results concerning the sampling distribution of x, the sample mean, is that mx m. This
result says that the x values from all possible random samples of size n center around
m, the population mean. For example, if m 100, the distribution is centered at 100,
whereas if m 5200, then the distribution is centered at 5200. Therefore, x is an
unbiased statistic for estimating m. Similarly, because the sampling distribution of p
is centered at p, it follows that p is an unbiased statistic for estimating a population
proportion.
Using an unbiased statistic that also has a small standard deviation ensures that
there will be no systematic tendency to under- or overestimate the value of the population characteristic and that estimates will almost always be relatively close to the
true value.
Given a choice between several unbiased statistics that could be used for estimating a population characteristic, the best statistic to use is the one with the smallest standard deviation.
Consider the problem of estimating a population mean, m. The obvious choice of
statistic for obtaining a point estimate of m is the sample mean, x, an unbiased statistic
for this purpose. However, when the population distribution is symmetric, x is not the
only choice. Other unbiased statistics for estimating m in this case include the sample
median and any trimmed mean (with the same number of observations trimmed from
each end of the ordered sample). Which statistic should be used? The following facts
may be helpful in making a choice.
1. If the population distribution is normal, then x has a smaller standard deviation than
any other unbiased statistic for estimating m. However, in this case, a trimmed mean
with a small trimming percentage (such as 10%) performs almost as well as x.
2. When the population distribution is symmetric with heavy tails compared to the
normal curve, a trimmed mean is a better statistic than x for estimating m.
When the population distribution is unquestionably normal, the choice is clear:
Use x to estimate m. However, with a heavy-tailed distribution, a trimmed mean gives
protection against one or two outliers in the sample that might otherwise have a large
effect on the value of the estimate.
Now consider estimating another population characteristic, the population variance s2. The sample variance
s2 g 1x x2 2
n1
is a good choice for obtaining a point estimate of the population variance s2. It can be
shown that s2 is an unbiased statistic for estimating s2; that is, whatever the value of
s2, the sampling distribution of s2 is centered at that value. It is precisely for this rea-
480
Chapter 9
■
Estimation Using a Single Sample
son—to obtain an unbiased statistic—that the divisor (n 1) is used. An alternative
statistic is the average squared deviation
g 1x x2 2
n
which one might think has a more natural divisor than s2. However, the average
squared deviation is biased, with its values tending to be smaller, on average, than s2.
..........................................................................................................................................
Example 9.3
Airborne Times for Flight 448
● The Bureau of Transportation Statistics provides data on U.S. airline flights. The
airborne times (in minutes) for United Airlines flight 448 from Albuquerque to Denver on 10 randomly selected days between January 1, 2003, and March 31, 2003, are
57
54
55
51
56
48
52
51
59
59
2
For these data gx 542, gx 29,498, n 10, and
1g x2 2
n
15422 2
29.498 10
121.6
2
2
a 1x x2 a x Let s2 denote the true variance in airborne time for flight 448. Using the sample
variance s2 to provide a point estimate of s2 yields
s2 g 1x x2 2
121.6
13.51
n1
9
Using the average squared deviation (with divisor n 10), the resulting point estimate is
g 1x x2 2
121.6
12.16
n
10
Because s2 is an unbiased statistic for estimating s2, most statisticians would recommend using the point estimate 13.51.
■
An obvious choice of a statistic for estimating the population standard deviation
s is the sample standard deviation s. For the data given in Example 9.3,
s 113.51 3.68
Unfortunately, the fact that s2 is an unbiased statistic for estimating s2 does not imply
that s is an unbiased statistic for estimating s. The sample standard deviation tends to
underestimate slightly the true value of s. However, unbiasedness is not the only criterion by which a statistic can be judged, and there are other good reasons for using
s to estimate s. In what follows, whenever we need to estimate s based on a single
random sample, we use the statistic s to obtain a point estimate.
● Data set available online
9.1
■ Exercises
■
Point Estimation 481
9 . 1 – 9 . 1 0 ...........................................................................................................
9.1 ▼ Three different statistics are being considered for
estimating a population characteristic. The sampling distributions of the three statistics are shown in the following
illustration:
2331
1852
Statistic III
Statistic I
True value of
population
characteristic
Which statistic would you recommend? Explain your
choice.
9.2 Why is an unbiased statistic generally preferred over
a biased statistic for estimating a population characteristic? Does unbiasedness alone guarantee that the estimate
will be close to the true value? Explain. Under what circumstances might you choose a biased statistic over an
unbiased statistic if two statistics are available for estimating a population characteristic?
9.3 Consumption of fast food is a topic of interest to researchers in the field of nutrition. The article “Effects of
Fast-Food Consumption on Energy Intake and Diet Quality Among Children” (Pediatrics [2004]: 112–118) reported that 1720 of those in a random sample of 6212
U.S. children indicated that on a typical day they ate fast
food. Estimate p, the proportion of children in the U.S.
who eat fast food on a typical day.
9.4 ● Data consistent with summary quantities in the
article referenced in Exercise 9.3 on total calorie consumption on a particular day are given for a sample of
children who did not eat fast food on that day and for a
sample of children who did eat fast food on that day. Assume that it is reasonable to regard these samples as representative of the population of children in the United
States.
1918
1777
1009
1765
1730
1827
1469
1648
2053
1506
2143
2669
1981
0000
934
875
2328
2207
2434
1811
2267
1250
2526
2117
1195
0000
Fast Food
2523
890
Statistic II
Bold exercises answered in back
No Fast Food
1758
1511
a. Use the given information to estimate the mean calorie
intake for children in the United States on a day when no
fast food is consumed.
b. Use the given information to estimate the mean calorie
intake for children in the United States on a day when fast
food is consumed.
c. Use the given information to estimate the produce estimates of the standard deviations of calorie intake for days
when no fast food is consumed and for days when fast
food is consumed.
9.5 Each person in a random sample of 20 students at a
particular university was asked whether he or she is registered to vote. The responses (R registered, N not registered) are given here:
R R N R N N R R R N R R R R R N R R R N
Use these data to estimate p, the true proportion of all students at the university who are registered to vote.
9.6 A study reported in Newsweek (December 23, 1991)
involved a sample of 935 smokers. Each individual received a nicotine patch, which delivers nicotine to the
bloodstream but at a much slower rate than cigarettes
do. Dosage was decreased to 0 over a 12-week period.
Suppose that 245 of the subjects were still not smoking
6 months after treatment (this figure is consistent with information given in the article). Estimate the percentage
of all smokers who, when given this treatment, would refrain from smoking for at least 6 months.
9.7 ● The article “Sensory and Mechanical Assessment
of the Quality of Frankfurters” (Journal of Texture Studies
[1990]: 395–409) reported the following salt content (percentage by weight) for 10 frankfurters:
2.26 2.11 1.64 1.17 1.64 2.36 1.70 2.10 2.19 2.40
a. Use the given data to produce a point estimate of m, the
true mean salt content for frankfurters.
● Data set available online but not required
▼ Video solution available
482
Chapter 9
■
Estimation Using a Single Sample
b. Use the given data to produce a point estimate of s2,
the variance of salt content for frankfurters.
c. Use the given data to produce an estimate of s, the
standard deviation of salt content. Is the statistic you used
to produce your estimate unbiased?
9.8 ● The following data on gross efficiency (ratio of
work accomplished per minute to calorie expenditure per
minute) for trained endurance cyclists were given in the
article “Cycling Efficiency Is Related to the Percentage of
Type I Muscle Fibers” (Medicine and Science in Sports
and Exercise [1992]: 782–88):
18.3 18.9 19.0 20.9 21.4 20.5 20.1 20.1
20.8 20.5 19.9 20.5 20.6 22.1 21.9 21.2
20.5 22.6 22.6 20.5 20.6 22.1 21.9 21.2
a. Assuming that the distribution of gross energy in the
population of all endurance cyclists is normal, give a point
estimate of m, the population mean gross efficiency.
b. Making no assumptions about the shape of the population distribution, estimate the proportion of all such cyclists whose gross efficiency is at most 20.
9.9 ● A random sample of n 12 four-year-old red pine
trees was selected, and the diameter (in inches) of each
tree’s main stem was measured. The resulting observations
are as follows:
11.3 10.7 12.4 15.2 10.1 12.1 16.2 10.5
11.4 11.0 10.7 12.0 10.1 12.1 16.2 10.5
a. Compute a point estimate of s, the population standard
deviation of main stem diameter. What statistic did you
use to obtain your estimate?
b. Making no assumptions about the shape of the population distribution of diameters, give a point estimate for the
Bold exercises answered in back
population median diameter. What statistic did you use to
obtain the estimate?
c. Suppose that the population distribution of diameter is
symmetric but with heavier tails than the normal distribution. Give a point estimate of the population mean diameter based on a statistic that gives some protection against
the presence of outliers in the sample. What statistic did
you use?
d. Suppose that the diameter distribution is normal. Then
the 90th percentile of the diameter distribution is m 1.28s (so 90% of all trees have diameters less than this
value). Compute a point estimate for this percentile. (Hint:
First compute an estimate of m in this case; then use it
along with your estimate of s from Part (a).)
9.10 ● A random sample of 10 houses in a particular
area, each of which is heated with natural gas, is selected,
and the amount of gas (in therms) used during the month
of January is determined for each house. The resulting
observations are as follows:
103 156 118 89 125 147 122 109 138 99
a. Let mJ denote the average gas usage during January
by all houses in this area. Compute a point estimate of mJ.
b. Suppose that 10,000 houses in this area use natural gas
for heating. Let t denote the total amount of gas used by
all of these houses during January. Estimate t using the
data of Part (a). What statistic did you use in computing
your estimate?
c. Use the data in Part (a) to estimate p, the proportion
of all houses that used at least 100 therms.
d. Give a point estimate of the population median usage based on the sample of Part (a). Which statistic did
you use?
● Data set available online but not required
▼ Video solution available
........................................................................................................................................
9.2
Large-Sample Confidence Interval for a
Population Proportion
In Section 9.1 we saw how to use a statistic to produce a point estimate of a population characteristic. The value of a point estimate depends on which sample, out of all
the possible samples, happens to be selected. Different samples usually yield different
estimates as a result of chance differences from one sample to another. Because of
sampling variability, rarely is the point estimate from a sample exactly equal to the
true value of the population characteristic. We hope that the chosen statistic produces
an estimate that is close, on average, to the true value. Although a point estimate may
9.2
■
Large-Sample Confidence Interval for a Population Proportion 483
represent our best single-number guess for the value of the population characteristic,
it is not the only plausible value. These considerations suggest the need to indicate in
some way a range of plausible values for the population characteristic. A point estimate by itself does not provide this information.
As an alternative to a point estimate, we can report an interval of reasonable values for the population characteristic based on the sample data. For example, we might
be confident that for all calls made from AT&T pay phones, the proportion p of calls
that are billed to a credit card is in the interval from .53 to .57. The narrowness of this
interval implies that we have rather precise information about the value of p. If, with
the same high degree of confidence, we could only state that p was between .32 and
.74, it would be clear that we had relatively imprecise knowledge of the value of p.
D
E F I N I T I O N
A confidence interval (CI) for a population characteristic is an interval of
plausible values for the characteristic. It is constructed so that, with a chosen degree of confidence, the value of the characteristic will be captured between the lower and upper endpoints of the interval.
Associated with each confidence interval is a confidence level. The confidence
level provides information on how much “confidence” we can have in the method used
to construct the interval estimate (not our confidence in any one particular interval).
Usual choices for confidence levels are 90%, 95%, and 99%, although other levels are
also possible. If we were to construct a 95% confidence interval using the technique
to be described shortly, we would be using a method that is “successful” 95% of the
time. That is, if this method was used to generate an interval estimate over and over
again with different samples, in the long run 95% of the resulting intervals would capture the true value of the characteristic being estimated. Similarly, a 99% confidence
interval is one that is constructed using a method that is, in the long run, successful in
capturing the true value of the population characteristic 99% of the time.
D
E F I N I T I O N
The confidence level associated with a confidence interval estimate is the
success rate of the method used to construct the interval.
Many factors influence the choice of confidence level. These factors will be discussed after we develop the method for constructing confidence intervals. We first consider a large-sample confidence interval for a population proportion p.
Often an investigator wishes to make an inference about the proportion of individuals or objects in a population that possess a particular property of interest. For
example, a university administrator might be interested in the proportion of students
who prefer a new web-based computer registration system to the previous registration
method. In a different setting, a quality control engineer might be concerned about the
proportion of defective parts manufactured using a particular process.
484
Chapter 9
■
Estimation Using a Single Sample
Let p be the proportion of the population that possess the property of interest. Previously, we used the sample proportion
p
number in the sample that possess the property of interest
n
to calculate a point estimate of p. We can also use p to form a confidence interval
for p.
Although a small-sample confidence interval for p can be obtained, our focus is
on the large-sample case. The justification for the large-sample interval rests on properties of the sampling distribution of the statistic p:
1. The sampling distribution of p is centered at p; that is, mp p. Therefore, p is an
unbiased statistic for estimating p.
p11 p 2
2. The standard deviation of p is sp n
B
3. As long as n is large (np 10 and n(1 p) 10) and the sample size is less than
10% of the population size, the sampling distribution of p is well approximated by
a normal curve.
The accompanying box summarizes these properties.
When n is large, the statistic p has a sampling distribution that is approximately normal with
p11 p2
.
mean p and standard deviation
n
B
The development of a confidence interval for p is easier to follow if we select a
particular confidence level. For a confidence level of 95%, Appendix Table 2, the table
of standard normal (z) curve areas, can be used to determine a value z* such that a central area of .95 falls between z* and z*. In this case, the remaining area of .05 is divided equally between the two tails, as shown in Figure 9.2. The total area to the left
of the desired z* is .975 (.95 central area .025 area below z*). By locating .9750
in the body of Appendix Table 2, we find that the corresponding z critical value is 1.96.
Capturing a
central area of .95 under the
z curve.
Figure 9.2
Central area .95
Upper-tail
area .025
Lower-tail
area .025
z* 1.96
z* 1.96
Generalizing this result to normal distributions other than the standard normal distribution tells us that for any normal distribution, about 95% of the values are within
1.96 standard deviations of the mean. Because (for large random samples) the sampling distribution of p is approximately normal with mean mp p and standard devip11 p 2
, we get the following result.
ation sp n
B
9.2
■
Large-Sample Confidence Interval for a Population Proportion 485
When n is large, approximately 95% of all samples of size n will result in a value of p that is
p11 p 2
within 1.96sp 1.96
of the true population proportion p.
n
B
If p is within 1.96
B
p 1.96
B
p11 p2
of p, this means the interval
n
p11 p 2
n
to p 1.96
B
p11 p2
n
will capture p (and this will happen for 95% of all possible samples). However, if p is
p11 p 2
farther away from p than 1.96
(which will happen for about 5% of all
n
B
possible samples), the interval will not include the true value of p. This is shown in
Figure 9.3.
The population proportion p is captured in the interval from
p11 p2
p 1.96
to
n
B
Approximate
sampling distribution
of p
Figure 9.3
π
p11 p2
p 1.96
when
n
B
p11 p 2
n
B
p is within 1.96
1.96
π(1 π) 1.96
–––––––
n
π(1 π)
–––––––
n
of p.
p 1.96
π(1 π)
––––––_–
n
interval
captures π
)
(
p
p 1.96
π(1 π)
––––––_–
n
)
(
p 1.96
π(1 π)
––––––_–
n
p
p 1.96
π(1 π)
––––––_–
n
interval does
not capture π
Because p is within 1.96sp of p 95% of the time, this implies that in repeated
sampling, 95% of the time the interval
p11 p 2
n
B
p 1.96
to p 1.96
B
p11 p2
n
will contain p.
p11 p2
must be estimated. As long as the sample size
n
B
p11 p2
p11 p2
.
is large, the value of
can be used in place of
n
n
B
B
Since p is unknown,
486
Chapter 9
■
Estimation Using a Single Sample
When n is large, a 95% confidence interval for p is
p11 p2
p11 p2
, p 1.96
b
n
n
B
B
a p 1.96
An abbreviated formula for the interval is
p11 p 2
n
B
p 1.96
p11 p2
p11 p 2
gives the upper endpoint of the interval and p 1.96
n
n
B
B
where p 1.96
gives the lower endpoint of the interval.
The interval can be used as long as
1. np 10 and n(1 p) 10,
2. the sample size is less than 10% of the population size if sampling is without replacement,
3. the sample can be regarded as a random sample from the population of interest.
..........................................................................................................................................
Example 9.4
Affirmative Action Continued
Let’s return to the information from the survey on attitudes toward affirmative action
(see the chapter introduction and Example 9.1):
Total number of people surveyed: 1013
Number who believe affirmative action programs should be continued: 537
A point estimate of p, the true proportion of U.S. adults who believe that affirmative
action programs should be continued, is
p
537
.530
1013
Because np (1013)(.53) 537 10 and n(1 p) (1013)(.47) 476 10
and the adults in the sample were randomly selected from a large population, the
large-sample interval can be used. For a 95% confidence level, a confidence interval
for p is
p 1.96
B
1.530 2 1.470 2
p11 p2
5.30 1.96
n
1013
B
.530 11.962 1.016 2
.530 .031
1.499, .5612
Based on this sample, we can be 95% confident that p, the true proportion who believe that affirmative action programs should continue, is between .499 and .561. We
9.2
■
Large-Sample Confidence Interval for a Population Proportion 487
used a method to construct this estimate that in the long run will successfully capture
the true value of p 95% of the time.
■
π
The 95% confidence interval for p calculated in Example
9.4 is (.499, .561). It is tempting to say that there is a “probability” of .95 that p is between .499 and .561. Do not yield to this
temptation! The 95% refers to the percentage of all possible
samples resulting in an interval that includes p. In other words,
if we take sample after sample from the population and use each
one separately to compute a 95% confidence interval, in the long
run roughly 95% of these intervals will capture p. Figure 9.4
illustrates this concept for intervals generated from 100 different random samples; 93 of the intervals include p, whereas 7 do
not. Any specific interval, and our interval (.499, .561) in particular, either includes p or it does not (remember, the value of
p is fixed but not known to us). We cannot make a chance (probability) statement concerning this particular interval. The confidence level 95% refers to the method used to construct the interval rather than to any particular interval, such as the one we
obtained.
The formula given for a 95% confidence interval can easily
be adapted for other confidence levels. The choice of a 95% confidence level led to the use of the z value 1.96 (chosen to capture a central area of .95 under the standard normal curve) in the
formula. Any other confidence level can be obtained by using
an appropriate z critical value in place of 1.96. For example,
suppose that we wanted to achieve a confidence level of 99%.
To obtain a central area of .99, the approximate z critical value
would have a cumulative area (area to the left) of .995, as illustrated in Figure 9.5. From Appendix Table 2, we find that the
corresponding z critical value is 2.58. A 99% confidence interval for p is then obtained by using 2.58 in place of 1.96 in the
formula for the 95% confidence interval.
*
*
*
*
*
*
*
π
One hundred 95% confidence intervals for p computed from 100 different random
samples (asterisks identify intervals that do not
include p).
Figure 9.4
Finding the z
critical value for a 99% confidence level.
Figure 9.5
upper-tail area
lower-tail area
.01
.005
2
.01
.005
2
central area
.99
z*
z*
Cumulative area .995
488
Chapter 9
■
Estimation Using a Single Sample
The Large-Sample Confidence Interval for P
The general formula for a confidence interval for a population proportion p when
1. p is the sample proportion from a random sample, and
2. the sample size n is large (np 10 and n(1 p) 10), and
3. if the sample is selected without replacement, the sample size is small relative to the population size (n is at most 10% of the population size)
is
p 1z critical value2
p11 p2
n
B
The desired confidence level determines which z critical value is used. The three most commonly used confidence levels, 90%, 95%, and 99%, use z critical values 1.645, 1.96, and 2.58,
respectively.
Note: This interval is not appropriate for small samples. It is possible to construct a confidence
interval in the small-sample case, but this is beyond the scope of this textbook.
Why settle for 95% confidence when 99% confidence is possible? Because the
higher confidence level comes with a price tag. The resulting interval is wider than the
p11 p2
b whereas the 99%
95% interval. The width of the 95% interval is 2 a 1.96
n
B
,
p11 p2
b . The higher reliability of the 99% interval
interval has width 2 a 2.58
n
B
(where “reliability” is specified by the confidence level) entails a loss in precision (as
indicated by the wider interval). In the opinion of many investigators, a 95% interval
is a reasonable compromise between reliability and precision.
..........................................................................................................................................
Example 9.5
Dangerous Driving
The article “Nine Out of Ten Drivers Admit in Survey to Having Done Something
Dangerous” (Knight Ridder Newspapers, July 8, 2005) reported the results of a survey of 1100 drivers. Of those surveyed, 990 admitted to careless or aggressive driving during the previous six months. Assuming that it is reasonable to regard this
sample of 1100 as representative of the population of drivers, we can use this information to construct an estimate of p, the true proportion of drivers who have engaged in careless or aggressive driving in the past six months.
For this sample
p
990
.900
1100
Because np 990 and n(1 p) 110 are both greater than or equal to 10, the
sample size is large enough to use the formula for a large-sample confidence interval. A 90% confidence interval for p is then
Step-by-step technology instructions available online
9.2
■
Large-Sample Confidence Interval for a Population Proportion 489
© The Image Bank / Wilfried Krecichwest /Getty Images
p 1z critical value2
p11 p2
1.9002 1.1002
.900 1.645
n
B
1100
B
.900 11.6452 1.0092
.900 .015
1.885, .915 2
Based on these sample data, we can be 90% confident that the true proportion of
drivers who have engaged in careless or aggressive driving in the past six months is
between .885 and .915. We have used a method to construct this interval estimate
that has a 10% error rate.
■
The confidence level for the z confidence interval for a population proportion is
only approximate. That is, when we report a 95% confidence interval for a population
proportion, the 95% confidence level implies that we have used a method that produces
an interval that includes the actual value of the population proportion 95% of the time
in repeated sampling. In fact, because the normal distribution is only an approximation to the sampling distribution of p, the true confidence level may differ somewhat
from the reported value. If the conditions (1) np 10 and n(1 p) 10 and (2) n is
at most 10% of the population size if sampling without replacement are met, the normal approximation is reasonable and the actual confidence level is usually quite close
to the reported level; this is why it is important to check these conditions before computing and reporting a z confidence interval for a population proportion.
What should you do if these conditions are not met? If the sample size is too small
to satisfy the np and n(1 p) 10 condition, an alternative procedure can be used.
onsult a statistician or a more advanced textbook in this case. If the condition that the
sample size is less than 10% of the population size when sampling without replacement is not satisfied, the z confidence interval tends to be conservative (i.e., it tends to
be wider than is necessary to achieve the desired confidence level). In this case, a finite
population correction factor can be used to obtain a more precise interval. Again, it
would be wise to consult a statistician or a more advanced textbook.
■
An Alternative to the Large-Sample z Interval ..............................................
Investigators have shown that in some instances, even when the sample size conditions
of the large-sample z confidence interval for a population proportion are met, the actual confidence level associated with the method may be noticeably different from the
reported confidence level. A modified interval that has an actual confidence level that
is closer to the reported confidence level is based on a modified sample proportion,
pmod, the proportion of successes after adding two successes and two failures to the
sample. Then pmod is
pmod number of successes 2
n4
pmod is used in place of p in the usual confidence interval formula. Properties of
this modified confidence interval are investigated in Activity 9.2 at the end of the
chapter.
490
Chapter 9
■
■
Estimation Using a Single Sample
General Form of a Confidence Interval ........................................................
Many confidence intervals have the same general form as the large-sample z interval
for p just considered. We started with a statistic p, from which a point estimate for p
was obtained. The standard deviation of this statistic is 2p11 p2/n. This resulted
in a confidence interval of the form
a
point estimate using
standard deviation
b 1critical value2 a
b
a specified statistic
of the statistic
Because p was unknown, we estimated the standard deviation of the statistic by
2p11 p 2/n, which yielded the interval
estimated
point estimate using
a
b 1critical value2 ° standard deviation ¢
a specified statistic
of the statistic
For a population characteristic other than p, a statistic for estimating the characteristic is selected. Then (drawing on statistical theory) a formula for the standard
deviation of the statistic is given. In practice, it is almost always necessary to estimate this standard deviation (using something analogous to 2p11 p 2/n rather than
2p11 p2/n, for example), so that the interval
estimated
point estimate using
a
b 1critical value2 ° standard deviation ¢
a specified statistic
of the statistic
is the prototype confidence interval. It is common practice to refer to both the standard
deviation of a statistic and the estimated standard deviation of a statistic as the standard error. In this textbook, when we use the term standard error, we mean the estimated standard deviation of a statistic.
D
E F I N I T I O N
The standard error of a statistic is the estimated standard deviation of the
statistic.
The 95% confidence interval for p is based on the fact that, for approximately
p11 p2
95% of all random samples, p is within 1.96
of p. The quantity
n
B
p11 p2
is sometimes called the bound on the error of estimation associated
1.96
n
B
with a 95% confidence level—we have 95% confidence that the point estimate p is no
farther than this quantity from p.
9.2
D
■
Large-Sample Confidence Interval for a Population Proportion 491
E F I N I T I O N
If the sampling distribution of a statistic is (at least approximately) normal,
the bound on error of estimation, B, associated with a 95% confidence interval is (1.96) # (standard error of the statistic).
■
Choosing the Sample Size ...............................................................................
Before collecting any data, an investigator may wish to determine a sample size for
which a particular value of the bound on the error is achieved. For example, with p
representing the true proportion of students at a university who purchase textbooks
over the Internet, the objective of an investigation may be to estimate p to within .05
with 95% confidence. The value of n necessary to achieve this is obtained by equating
p11 p 2
.05 to 1.96
and solving for n.
n
B
In general, suppose that we wish to estimate p to within an amount B (the specified bound on the error of estimation) with 95% confidence. Finding the necessary
sample size requires solving the equation
p11 p2
n
B
B 1.96
Solving this equation for n results in
n p11 p 2 a
1.96 2
b
B
Unfortunately, the use of this formula requires the value of p, which is unknown. One
possible way to proceed is to carry out a preliminary study and use the resulting data
to get a rough estimate of p. In other cases, prior knowledge may suggest a reasonable
estimate of p. If there is no reasonable basis for estimating p and a preliminary study
is not feasible, a conservative solution follows from the observation that p(1 p) is
never larger than .25 (its value when p .5). Replacing p(1 p) with .25, the maximum value, yields
n .25 a
1.96 2
b
B
Using this formula to obtain n gives us a sample size for which we can be 95% confident that p will be within B of p, no matter what the value of p.
The sample size required to estimate a population proportion p to within an amount B with
95% confidence is
n p11 p2 a
1.96 2
b
B
The value of p may be estimated using prior information. In the absence of any such information, using p .5 in this formula gives a conservatively large value for the required sample
size (this value of p gives a larger n than would any other value).
Chapter 9
492
■
Estimation Using a Single Sample
..........................................................................................................................................
Example 9.6
Sniffing Out Cancer
Researchers have found biochemical markers of cancer in the exhaled breath of cancer patients, but chemical analysis of breath specimens has not yet proven effective
in clinical diagnosis. The authors of the paper “Diagnostic Accuracy of Canine Scent
Detection in Early- and Late-Stage Lung and Breast Cancers” (Integrative Cancer
Therapies [2006]: 1–10) describe a study to investigate whether dogs can be trained
to identify the presence or absence of cancer by sniffing breath specimens. Suppose
we want to collect data that would allow us to estimate the long-run proportion of
accurate identifications for a particular dog that has completed training. The dog has
been trained to lie down when presented with a breath specimen from a cancer patient and to remain standing when presented with a specimen from a person who
does not have cancer. How many different breath specimens should be used if we
want to estimate the long-run proportion of correct identifications for this dog to
within .05 with 95% confidence?
Using a conservative value of p .5 in the formula for required sample size
gives
n p11 p2 a
1.96 2
1.96 2
b .25 a
b 384.16
B
.05
Thus, a sample of at least 385 breath specimens should be used. Note that in sample
size calculations, we always round up.
■
■
E x e r c i s e s 9 . 1 1 – 9 . 2 9 ..........................................................................................................
9.11 ▼ For each of the following choices, explain which
would result in a wider large-sample confidence interval
for p:
a. 90% confidence level or 95% confidence level
b. n 100 or n 400
9.12 The formula used to compute a large-sample confidence interval for p is
p 1z critical value 2
p11 p 2
n
B
What is the appropriate z critical value for each of the following confidence levels?
a. 95%
d. 80%
b. 90%
e. 85%
c. 99%
Bold exercises answered in back
9.13 The use of the interval
p 1z critical value 2
p11 p 2
n
B
requires a large sample. For each of the following combinations of n and p, indicate whether the given interval
would be appropriate.
a. n 50 and p .30
b. n 50 and p .05
c. n 15 and p .45
d. n 100 and p .01
e. n 100 and p .70
f. n 40 and p .25
g. n 60 and p .25
h. n 80 and p .10
● Data set available online but not required
▼ Video solution available
9.2
■
Large-Sample Confidence Interval for a Population Proportion 493
9.14 Discuss how each of the following factors affects the
width of the confidence interval for p:
a. The confidence level
b. The sample size
c. The value of p
9.15 According to an AP-Ipsos poll (June 15, 2005), 42%
of 1001 randomly selected adult Americans made plans in
May 2005 based on a weather report that turned out to be
wrong.
a. Construct and interpret a 99% confidence interval for
the proportion of Americans who made plans in May 2005
based on an incorrect weather report.
b. Do you think it is reasonable to generalize this estimate
to other months of the year? Explain.
9.16 The article “Students Increasingly Turn to Credit
Cards” (San Luis Obispo Tribune, July 21, 2006) reported
that 37% of college freshmen and 48% of college seniors
carry a credit card balance from month to month. Suppose that the reported percentages were based on random samples of 1000 college freshmen and 1000 college
seniors.
a. Construct a 90% confidence interval for the proportion
of college freshmen who carry a credit card balance from
month to month.
b. Construct a 90% confidence interval for the proportion
of college seniors who carry a credit card balance from
month to month.
c. Explain why the two 90% confidence intervals from
Parts (a) and (b) are not the same width.
9.17 ▼ The article “CSI Effect Has Juries Wanting More
Evidence” (USA Today, August 5, 2004) examines how
the popularity of crime-scene investigation television
shows is influencing jurors’ expectations of what evidence
should be produced at a trial. In a survey of 500 potential
jurors, one study found that 350 were regular watchers of
at least one crime-scene forensics television series.
a. Assuming that it is reasonable to regard this sample of
500 potential jurors as representative of potential jurors in
the United States, use the given information to construct
and interpret a 95% confidence interval for the true proportion of potential jurors who regularly watch at least
one crime-scene investigation series.
b. Would a 99% confidence interval be wider or narrower
than the 95% confidence interval from Part (a)?
Bold exercises answered in back
9.18 In a survey of 1000 randomly selected adults in the
United States, participants were asked what their most favorite and what their least favorite subject was when they
were in school (Associated Press, August 17, 2005). In
what might seem like a contradiction, math was chosen
more often than any other subject in both categories! Math
was chosen by 230 of the 1000 as the favorite subject, and
it was also chosen by 370 of the 1000 as the least favorite
subject.
a. Construct a 95% confidence interval for the proportion
of U.S. adults for whom math was the favorite subject in
school.
b. Construct a 95% confidence interval for the proportion
of U.S. adults for whom math was the least favorite subject.
9.19 The report “2005 Electronic Monitoring & Surveillance Survey: Many Companies Monitoring, Recording,
Videotaping—and Firing—Employees” (American Management Association, 2005) summarized the results of a
survey of 526 U.S. businesses. The report stated that 137
of the 526 businesses had fired workers for misuse of the
Internet and 131 had fired workers for email misuse. For
purposes of this exercise, assume that it is reasonable to
regard this sample as representative of businesses in the
United States.
a. Construct and interpret a 95% confidence interval for
the proportion of U.S. businesses that have fired workers
for misuse of the Internet.
b. What are two reasons why a 90% confidence interval
for the proportion of U.S. businesses that have fired workers for misuse of email would be narrower than the 95%
confidence interval computed in Part (a).
9.20 In an AP-AOL sports poll (Associated Press, December 18, 2005), 394 of 1000 randomly selected U.S. adults
indicated that they considered themselves to be baseball
fans. Of the 394 baseball fans, 272 stated that they thought
the designated hitter rule should either be expanded to
both baseball leagues or eliminated.
a. Construct a 95% confidence interval for the proportion of U.S. adults that consider themselves to be baseball fans.
b. Construct a 95% confidence interval for the proportion
of those who consider themselves to be baseball fans that
think the designated hitter rule should be expanded to both
leagues or eliminated.
c. Explain why the confidence intervals of Parts (a) and
● Data set available online but not required
▼ Video solution available
494
Chapter 9
■
Estimation Using a Single Sample
(b) are not the same width even though they both have a
confidence level of 95%.
9.21 The article “Viewers Speak Out Against Reality TV”
(Associated Press, September 12, 2005) included the following statement: “Few people believe there’s much reality in reality TV: a total of 82 percent said the shows are
either ‘totally made up’ or ‘mostly distorted’.” This statement was based on a survey of 1002 randomly selected
adults. Compute and interpret a bound on the error of estimation for the reported percentage.
9.22 One thousand randomly selected adult Americans
participated in a survey conducted by the Associated Press
(June, 2006). When asked “Do you think it is sometimes
justified to lie or do you think lying is never justified?”
52% responded that lying was never justified. When asked
about lying to avoid hurting someone’s feelings, 650 responded that this was often or sometimes OK.
a. Construct a 90% confidence interval for the proportion
of adult Americans who think lying is never justified.
b. Construct a 90% confidence interval for the proportion
of adult American who think that it is often or sometimes
OK to lie to avoid hurting someone’s feelings.
c. Based on the confidence intervals from Parts (a) and
(b), comment on the apparent inconsistency in the responses given by the individuals in this sample.
ferent countries (San Luis Obispo Tribune, November 21,
2002). The society found that 10% of the participants
could not identify their own country on a blank world map.
a. Construct a 90% confidence interval for the proportion
who can identify their own country on a blank world map.
b. What assumptions are necessary for the confidence interval in Part (a) to be valid?
c. To what population would it be reasonable to generalize the confidence interval estimate from Part (a)?
9.25 ▼ “Tongue Piercing May Speed Tooth Loss, Researchers Say” is the headline of an article that appeared
in the San Luis Obispo Tribune (June 5, 2002). The article
describes a study of 52 young adults with pierced tongues.
The researchers found receding gums, which can lead to
tooth loss, in 18 of the participants. Construct a 95% confidence interval for the proportion of young adults with
pierced tongues who have receding gums. What assumptions must be made for use of the z confidence interval to
be appropriate?
9.26 USA Today (October 14, 2002) reported that 36% of
adult drivers admit that they often or sometimes talk on a
cell phone when driving. This estimate was based on data
from a sample of 1004 adult drivers, and a bound on the
error of estimation of 3.1% was reported. Explain how the
given bound on the error can be justified.
9.23 The article “Doctors Cite Burnout in Mistakes” (San
Luis Obispo Tribune, March 5, 2002) reported that many
doctors who are completing their residency have financial
struggles that could interfere with training. In a sample of
115 residents, 38 reported that they worked moonlighting
jobs and 22 reported a credit card debt of more than $3000.
Suppose that it is reasonable to consider this sample of 115
as a random sample of all medical residents in the United
States.
a. Construct and interpret a 95% confidence interval for
the proportion of U.S. medical residents who work moonlighting jobs.
b. Construct and interpret a 90% confidence interval for
the proportion of U.S. medical residents who have a credit
card debt of more than $3000.
c. Give two reasons why the confidence interval in Part (a)
is wider than the confidence interval in Part (b).
9.28 In a study of 1710 schoolchildren in Australia (Herald Sun, October 27, 1994), 1060 children indicated that
they normally watch TV before school in the morning.
(Interestingly, only 35% of the parents said their children
watched TV before school!) Construct a 95% confidence
interval for the true proportion of Australian children who
say they watch TV before school. What assumption about
the sample must be true for the method used to construct
the interval to be valid?
9.24 The National Geographic Society conducted a study
that included 3000 respondents, age 18 to 24, in nine dif-
9.29 ▼ A consumer group is interested in estimating the
proportion of packages of ground beef sold at a particular
Bold exercises answered in back
9.27 The Gallup Organization conducts an annual survey
on crime. It was reported that 25% of all households experienced some sort of crime during the past year. This estimate was based on a sample of 1002 randomly selected
adults. The report states, “One can say with 95% confidence that the margin of sampling error is 3 percentage points.” Explain how this statement can be justified.
● Data set available online but not required
▼ Video solution available
9.3
store that have an actual fat content exceeding the fat content stated on the label. How many packages of ground
Bold exercises answered in back
■
Confidence Interval for a Population Mean 495
beef should be tested to estimate this proportion to within
.05 with 95% confidence?
● Data set available online but not required
▼ Video solution available
........................................................................................................................................
9.3
Confidence Interval for a Population Mean
In this section, we consider how to use information from a random sample to construct
a confidence interval estimate of a population mean, m. We begin by considering the
case in which (1) s, the population standard deviation, is known (not realistic, but
we will see shortly how to handle the more realistic situation where s is unknown)
and (2) the sample size n is large enough for the Central Limit Theorem to apply. In
this case, the following three properties about the sampling distribution of x hold:
1. The sampling distribution of x is centered at m, so x is an unbiased statistic for estimating m 1mx m2.
s
.
2. The standard deviation of x is sx 1n
3. As long as n is large (generally n 30), the sampling distribution of x is approximately normal, even when the population distribution itself is not normal.
The same reasoning that was used to develop the large-sample confidence interval for a population proportion p can be used to obtain a confidence interval estimate
for m.
The One-Sample z Confidence Interval for M
The general formula for a confidence interval for a population mean m when
1. x is the sample mean from a random sample,
2. the sample size n is large (generally n 30), and
3. s, the population standard deviation, is known
is
x 1z critical value2 a
s
b
1n
..........................................................................................................................................
Example 9.7
Cosmic Radiation
Cosmic radiation levels rise with increasing altitude, prompting researchers to consider how pilots and flight crews might be affected by increased exposure to cosmic
radiation. The paper “Estimated Cosmic Radiation Doses for Flight Personnel”
(Space Medicine and Medical Engineering [2002]: 265–269) reported a mean annual
cosmic radiation dose of 219 mrems for a sample of flight personnel of Xinjiang
Airlines. Suppose that this mean was based on a random sample of 100 flight crew
members.
496
Chapter 9
■
Estimation Using a Single Sample
Let m denote the true mean annual cosmic radiation exposure for Xinjiang Airlines flight crew members. Although s, the true population standard deviation, is not
usually known, suppose for illustrative purposes that s 35 mrem is known. Because the sample size is large and s is known, a 95% confidence interval for m is
x 1z critical value2 a
s
35
b 219 11.962 a
b
1n
1100
219 6.86
1212.14, 225.862
Based on this sample, plausible values of m, the true mean annual cosmic radiation exposure for Xinjaing Airlines flight crew members, are between 212.14 and
225.86 mrem. A 95% confidence level is associated with the method used to produce
this interval estimate.
■
The confidence interval just introduced is appropriate when s is known and n is
large, and it can be used regardless of the shape of the population distribution. This is
because this confidence interval is based on the Central Limit Theorem, which says
that when n is sufficiently large, the sampling distribution of x is approximately normal for any population distribution. When n is small, the Central Limit Theorem cannot be used to justify the normality of the x sampling distribution, so the z confidence
interval can not be used. One way to proceed in the small-sample case is to make a
specific assumption about the shape of the population distribution and then to use a
method of estimation that is valid only under this assumption.
The one instance where this is easy to do is when it is reasonable to believe that
the population distribution is normal in shape. Recall that for a normal population distribution the sampling distribution of x is normal even for small sample sizes. So, if n
is small but the population distribution is normal, the same confidence interval formula
just introduced can still be used.
If n is small (generally n 30) but it is reasonable to believe that the distribution of values in
the population is normal, a confidence interval for m (when s is known) is
x 1z critical value2 a
s
b
1n
There are several ways that sample data can be used to assess the plausibility of
normality. The two most common ways are to look at a normal probability plot of the
sample data (looking for a plot that is reasonably straight) and to construct a boxplot
of the data (looking for approximate symmetry and no outliers).
■
Confidence Interval for m When s Is Unknown .............................................
The confidence intervals just developed have an obvious drawback: To compute the interval endpoints, s must be known. Unfortunately, this is rarely the case in practice.
We now turn our attention to the situation when s is unknown. The development of
the confidence interval in this instance depends on the assumption that the population
9.3
■
Confidence Interval for a Population Mean 497
distribution is normal. This assumption is not critical if the sample size is large, but it
is important when the sample size is small.
To understand the derivation of this confidence interval, it is instructive to begin
by taking another look at the previous 95% confidence interval. We know that mx m
s
and sx . Also, when the population distribution is normal, the x distribution is
1n
normal. These facts imply that the standardized variable
z
xm
s
1n
has approximately a standard normal distribution. Because the interval from 1.96 to
1.96 captures an area of .95 under the z curve, approximately 95% of all samples result in an x value that satisfies
1.96 xm
1.96
s
1n
Manipulating these inequalities to isolate m in the middle results in the equivalent inequalities:
x 1.96 a
s
s
b m x 1.96 a
b
1n
1n
s
b is the lower endpoint of the 95% large-sample confidence
1n
s
b is the upper endpoint.
interval for m, and x 1.96 a
1n
If s is unknown, we must use the sample data to estimate s. If we use the sample
standard deviation as our estimate, the result is a different standardized variable denoted by t:
The term x 1.96 a
t
xm
s
1n
The value of s may not be all that close to s, especially when n is small. As a consequence, the use of s in place of s introduces extra variability, so the distribution of
t is more spread out than the standard normal (z) distribution. (The value of z varies
from sample to sample, because different samples generally result in different x values. There is even more variability in t, because different samples may result in different values of both x and s.)
To develop an appropriate confidence interval, we must investigate the probability distribution of the standardized variable t for a sample from a normal population.
This requires that we first learn about probability distributions called t distributions.
■
t Distributions ......................................................................................................
Just as there are many different normal distributions, there are also many different
t distributions. Although normal distributions are distinguished from one another by
498
Chapter 9
■
Estimation Using a Single Sample
their mean m and standard deviation s, t distributions are distinguished by a positive
whole number called the number of degrees of freedom (df). There is a t distribution
with 1 df, another with 2 df, and so on.
I m p o r t a n t P ro p e r t i e s o f t D i s t r i b u t i o n s
1. The t distribution corresponding to any fixed number of degrees of freedom is bell shaped
and centered at zero (just like the standard normal (z) distribution).
2. Each t distribution is more spread out than the standard normal (z) distribution.
3. As the number of degrees of freedom increases, the spread of the corresponding t distribution decreases.
4. As the number of degrees of freedom increases, the corresponding sequence of t distributions approaches the standard normal (z) distribution.
z curve
The properties discussed in the preceding box are illustrated in
Figure 9.6, which shows two t curves along with the z curve.
t curve for 4 df
Appendix Table 3 gives selected critical values for various
t distributions. The central areas for which values are tabulated
are .80, .90, .95, .98, .99, .998, and .999. To find a particular
critical value, go down the left margin of the table to the row
labeled with the desired number of degrees of freedom. Then
move over in that row to the column headed by the desired cen3 2
1
0
1
2
3
tral area. For example, the value in the 12-df row under the column corresponding to central area .95 is 2.18, so 95% of the
F i g u r e 9 . 6 Comparison of the z curve and
area under the t curve with 12 df lies between 2.18 and 2.18.
t curves for 12 df and 4 df.
Moving over two columns, we find the critical value for central
area .99 (still with 12 df) to be 3.06 (see Figure 9.7). Moving
down the .99 column to the 20-df row, we see the critical value is 2.85, so the area between 2.85 and 2.85 under the t curve with 20 df is .99.
t curve for 12 df
Figure 9.7
ues illustrated.
t critical val-
t curve
for 12 df
Shaded area .95
2.18
0
2.18
Shaded area .99
3.06
0
3.06
Notice that the critical values increase from left to right in each row of Appendix
Table 3. This makes sense because as we move to the right, we capture larger central
areas. In each column, the critical values decrease as we move downward, reflecting
decreasing spread for t distributions with larger degrees of freedom.
The larger the number of degrees of freedom, the more closely the t curve resembles the z curve. To emphasize this, we have included the z critical values as the last
row of the t table. Furthermore, once the number of degrees of freedom exceeds 30,
the critical values change little as the number of degrees of freedom increases. For this
reason, Appendix Table 3 jumps from 30 df to 40 df, then to 60 df, then to 120 df, and
9.3
■
Confidence Interval for a Population Mean 499
finally to the row of z critical values. If we need a critical value for a number of degrees of freedom between those tabulated, we just use the critical value for the closest
df. For df 120, we use the z critical values. Many graphing calculators calculate t
critical values for any number of degrees of freedom; so, if you are using such a calculator, it is not necessary to approximate the t critical values as described.
■
One-Sample t Confidence Interval ................................................................
xm
is approximately the z (standard
1s/ 1n 2
normal) distribution when n is large led to the z confidence interval when s is known.
In the same way, the following proposition provides the key to obtaining a confidence
interval when the population distribution is normal but s is unknown.
The fact that the sampling distribution of
Let x1, x2, ..., xn constitute a random sample from a normal population distribution. Then the
probability distribution of the standardized variable
t
xm
s
1n
is the t distribution with df n 1.
To see how this result leads to the desired confidence interval, consider the case
n 25. We use the t distribution with df 24 (n 1). From Appendix Table 3, the
interval between 2.06 and 2.06 captures a central area of .95 under the t curve with
24 df. This means that 95% of all samples (with n 25) from a normal population result in values of x and s for which
2.06 xm
2.06
s
1n
Algebraically manipulating these inequalities to isolate m yields
x 2.06 a
s
s
b m x 2.06 a
b
125
125
The 95% confidence interval for m in this situation extends from the lower endpoint
s
s
x 2.06 a
b to the upper endpoint x 2.06 a
b . This interval can be
125
125
written
x 2.06 a
s
b
125
The differences between this interval and the interval when s is known are the use
of the t critical value 2.06 rather than the z critical value 1.96 and the use of the sample
standard deviation as an estimate of s. The extra uncertainty that results from estimating s causes the t interval to be wider than the z interval.
500
Chapter 9
■
Estimation Using a Single Sample
If the sample size is something other than 25 or if the desired confidence level
is something other than 95%, a different t critical value (obtained from Appendix
Table 3) is used in place of 2.06.
The One-Sample t Confidence Interval for M
The general formula for a confidence interval for a population mean m based on a sample of
size n when
1. x is the sample mean from a random sample,
2. the population distribution is normal, or the sample size n is large (generally n 30),
and
3. s, the population standard deviation, is unknown
is
x 1t critical value2 a
s
b
1n
where the t critical value is based on df n 1. Appendix Table 3 gives critical values appropriate for each of the confidence levels 90%, 95%, and 99%, as well as several other less
frequently used confidence levels.
If n is large (generally n 30), the normality of the population distribution is
not critical. However, this confidence interval is appropriate for small n only when the
population distribution is (at least approximately) normal. If this is not the case, as
might be suggested by a normal probability plot or boxplot, another estimation method
should be used.
..........................................................................................................................................
Example 9.8
Waiting for Surgery
The Cardiac Care Network in Ontario, Canada collected information on the time between the date a patient was recommended for heart surgery and the surgery date for
cardiac patients in Ontario (“Wait Times Data Guide,” Ministry of Health and LongTerm Care, Ontario, Canada, 2006). The reported mean waiting time (in days) for
samples of patients for two cardiac procedures are given in the accompanying table.
(The standard deviations in the table were estimated from information on wait-time
variability included in the report.)
Surgical
Procedure
Sample Size
Mean
Wait Time
Standard Deviation
Bypass
Angiography
539
847
19
18
10
9
If we had access to the raw data (the 539 847 1386 individual wait time observations), we might begin by looking at boxplots. Data consistent with the given
9.3
■
Confidence Interval for a Population Mean 501
summary quantities were used to generate the boxplots of Figure 9.8. The boxplots for the two surgical procedures are similar. There are outliers in both data sets,
which might cause us to question the normality of the two wait-time distributions,
but because the sample sizes are large, it is still appropriate to use the t confidence
interval.
Figure 9.8
Boxplots for
Wait time (days)
Example 9.8.
60
*
*
*
*
*
*
*
*
Bypass
Angiography
50
40
30
20
10
0
As a next step, we can use the confidence interval of this section to estimate the true
mean wait time for each of the two procedures. Let’s first focus on the sample of bypass patients. For this group,
sample size n 539
sample mean wait time x 19
sample standard deviation s 10
The report referenced here indicated that it is reasonable to regard these data as representative of the Ontario population. So, with m denoting the mean wait time for
bypass surgery in Ontario, we can estimate m using a 90% confidence interval.
From Appendix Table 3, we use t critical value 1.645 (from the z critical value
row because df n 1 538 120, the largest number of degrees of freedom in
the table). The 90% confidence interval for m is
x 1t critical value2 a
s
10
b 19 11.6452 a
b
1n
1539
19 .709
118.291, 19.7092
Based on this sample, we are 90% confident that m is between 18.291 days and
19.709 days. This interval is fairly narrow indicating that our information about the
value of m is relatively precise.
A graphing calculator or any of the commercially available statistical computing packages can produce t confidence intervals. Confidence interval output from
MINITAB for the angiography data is shown here.
One-Sample T
N
Mean
847
18.0000
StDev
9.0000
SE Mean
0.3092
90% CI
(17.4908, 18.5092)
502
Chapter 9
■
Estimation Using a Single Sample
The 90% confidence interval for mean wait time for angiography extends from
17.4908 days to 18.5092 days. This interval is narrower than the 90% interval for
bypass surgery wait time for two reasons: the sample size is larger (847 rather than
539) and the sample standard deviation is smaller (9 rather than 10).
■
..........................................................................................................................................
Example 9.9
Selfish Chimps?
© Alan and Sandy Carey/Getty Images
●
The article “Chimps Aren’t Charitable” (Newsday, November 2, 2005) summarized the results of a research study published in the journal Nature. In this study,
chimpanzees learned to use an apparatus that dispensed food when either of two
ropes was pulled. When one of the ropes was pulled, only the chimp controlling the
apparatus received food. When the other rope was pulled, food was dispensed both
to the chimp controlling the apparatus and also to a chimp in the adjoining cage. The
accompanying data (approximated from a graph in the paper) represent the number
of times out of 36 trials that each of seven chimps chose the option that would provide food to both chimps (the “charitable” response).
23
22
21
24
19
20
20
Figure 9.9 is a normal probability plot of these data. The plot is reasonably straight,
so it seems plausible that the population distribution is approximately normal.
Normal Score
2
1
0
1
Normal
probability plot for data
of Example 9.9.
Figure 9.9
2
19
20
21
22
23
Number of Charitable Responses
24
Calculation of a confidence interval for the population mean number of charitable
responses requires x and s. From the given data, we compute
x 21.29
Step-by-step technology instructions available online
s 1.80
● Data set available online
9.3
■
Confidence Interval for a Population Mean 503
The t critical value for a 99% confidence interval based on 6 df is 3.71. The interval is
x 1t critical value2 a
s
1.80
b 21.29 13.712 a
b
1n
17
21.29 2.52
118.77, 23.812
A statistical software package could also have been used to compute the 99% confidence interval. The following is output from SPSS. The slight discrepancy between
the hand-calculated interval and the one reported by SPSS occurs because SPSS uses
more decimal accuracy in x, s, and t critical values.
One-Sample Statistics
CharitableResponses
N
7
Mean
21.2857
Std. Deviation
1.79947
Std. Error Mean
.68014
One-Sample
CharitableResponses
99% Confidence Interval of the Difference
Lower
Upper
18.7642
23.8073
With 99% confidence, we estimate the population mean number of charitable responses (out of 36 trials) to be between 18.77 and 23.81. Remember that the 99%
confidence level implies that if the same formula is used to calculate intervals for
sample after sample randomly selected from the population, in the long run 99% of
these intervals will capture m between the lower and upper confidence limits.
Notice that based on this interval, we would conclude that on average chimps
choose the charitable option more than half the time (18 out of 36 trials). The Newsday headline “Chimps Aren’t Charitable” was based on additional data from the
study indicating that chimps’ charitable behavior was no different when there was
another chimp in the adjacent cage than when the adjacent cage was empty. We will
revisit this study in Chapter 11 to investigate this further.
■
..........................................................................................................................................
© Stone/Mark Douet /Getty Images
E x a m p l e 9 . 1 0 Housework
How much time do school-age children spend helping with housework? The article
“The Three Corners of Domestic Labor: Mothers’, Fathers’, and Children’s Weekday
and Weekend Housework” (Journal of Marriage and the Family [1994]: 657–668)
gave information on the number of minutes per weekday spent on housework. The
following mean and standard deviation are for a random sample of 26 girls in twoparent families where both parents work full-time:
n 26
x 14.0
s 8.6
The authors of the article analyzed these data using methods designed for population distributions that are approximately normal. This assumption appears a bit
questionable based on the reported mean and standard deviation (it is impossible to
spend less than 0 min per day on housework, so the smallest possible value, 0, is
only 1.63 standard deviations below the mean). However, because the authors reported that there were no outliers in the data and because n is relatively close to 30,
504
Chapter 9
■
Estimation Using a Single Sample
we use the t confidence interval formula of this section to compute a 95% confidence interval.
Because n 26, df 25, and the appropriate t critical value is 2.06, the confidence interval is then
x 1t critical value2 a
s
8.6
b 14.0 12.062 a
b
1n
126
14.0 3.5
110.5, 17.5 2
Based on the sample data, we believe that the true mean time per weekday spent on
housework is between 10.5 and 17.5 min for girls in two-parent families where both
parents work. We used a method that has a 5% error rate to construct this interval.
We should be somewhat cautious in interpreting this confidence interval because of
the concern expressed about the normality of the population distribution.
■
■
Choosing the Sample Size ...............................................................................
When estimating m using a large sample or a small sample from a normal population
with known s, the bound B on the error of estimation associated with a 95% confidence interval is
B 1.96 a
s
b
1n
Before collecting any data, an investigator may wish to determine a sample size for
which a particular value of the bound is achieved. For example, with m representing
the average fuel efficiency (in miles per gallon, mpg) for all cars of a certain type, the
objective of an investigation may be to estimate m to within 1 mpg with 95% confidence. The value of n necessary to achieve this is obtained by setting B 1 and then
s
b for n.
solving 1 1.96 a
1n
In general, suppose that we wish to estimate m to within an amount B (the specified bound on the error of estimation) with 95% confidence. Finding the necessary
s
b for n. The result is
sample size requires solving the equation B 1.96 a
1n
n a
1.96s 2
b
B
Notice that, in general, a large value of s forces n to be large, as does a small
value of B.
Use of the sample-size formula requires that s be known, but this is rarely the
case in practice. One possible strategy for estimating s is to carry out a preliminary
study and use the resulting sample standard deviation (or a somewhat larger value, to
be conservative) to determine n for the main part of the study. Another possibility is
simply to make an educated guess about the value of s and to use that value to calculate n. For a population distribution that is not too skewed, dividing the range (the difference between the largest and the smallest values) by 4 often gives a rough idea of
the value of the standard deviation.
9.3
■
Confidence Interval for a Population Mean 505
The sample size required to estimate a population mean m to within an amount B with 95%
confidence is
n a
1.96s 2
b
B
If s is unknown, it may be estimated based on previous information or, for a population that is
not too skewed, by using (range)/4.
If the desired confidence level is something other than 95%, 1.96 is replaced by the appropriate z critical value (e.g., 2.58 for 99% confidence).
..........................................................................................................................................
E x a m p l e 9 . 1 1 Cost of Textbooks
The financial aid office wishes to estimate the mean cost of textbooks per quarter for
students at a particular university. For the estimate to be useful, it should be within
$20 of the true population mean. How large a sample should be used to be 95% confident of achieving this level of accuracy?
To determine the required sample size, we must have a value for s. The financial aid office is pretty sure that the amount spent on books varies widely, with most
values between $50 and $450. A reasonable estimate of s is then
range
450 50
400
100
4
4
4
The required sample size is
n a
11.962 11002 2
1.96s 2
b a
b 19.82 2 96.04
B
20
Rounding up, a sample size of 97 or larger is recommended.
■
■
E x e r c i s e s 9.30–9.50 ..............................................................................................................
9.30 Given a variable that has a t distribution with the
specified degrees of freedom, what percentage of the time
will its value fall in the indicated region?
a. 10 df, between 1.81 and 1.81
b. 10 df, between 2.23 and 2.23
c. 24 df, between 2.06 and 2.06
d. 24 df, between 2.80 and 2.80
e. 24 df, outside the interval from 2.80 to 2.80
f. 24 df, to the right of 2.80
g. 10 df, to the left of 1.81
Bold exercises answered in back
9.31 The formula used to compute a confidence interval
for the mean of a normal population when n is small is
x 1t critical value 2
s
1n
What is the appropriate t critical value for each of the following confidence levels and sample sizes?
a. 95% confidence, n 17
b. 90% confidence, n 12
c. 99% confidence, n 24
● Data set available online but not required
▼ Video solution available
506
Chapter 9
■
Estimation Using a Single Sample
d. 90% confidence, n 25
e. 90% confidence, n 13
f. 95% confidence, n 10
9.32 The two intervals (114.4, 115.6) and (114.1, 115.9)
are confidence intervals for m true average resonance
frequency (in hertz) for all tennis rackets of a certain type.
a. What is the value of the sample mean resonance
frequency?
b. The confidence level for one of these intervals is 90%
and for the other it is 99%. Which is which, and how can
you tell?
9.33 ▼ Samples of two different types of automobiles were
selected, and the actual speed for each car was determined
when the speedometer registered 50 mph. The resulting
95% confidence intervals for true average actual speed
were (51.3, 52.7) and (49.4, 50.6). Assuming that the two
sample standard deviations are identical, which confidence
interval is based on the larger sample size? Explain your
reasoning.
9.34 Suppose that a random sample of 50 bottles of a particular brand of cough medicine is selected and the alcohol content of each bottle is determined. Let m denote the
average alcohol content for the population of all bottles of
the brand under study. Suppose that the sample of 50 results in a 95% confidence interval for m of (7.8, 9.4).
a. Would a 90% confidence interval have been narrower
or wider than the given interval? Explain your answer.
b. Consider the following statement: There is a 95%
chance that m is between 7.8 and 9.4. Is this statement
correct? Why or why not?
c. Consider the following statement: If the process of selecting a sample of size 50 and then computing the corresponding 95% confidence interval is repeated 100 times,
95 of the resulting intervals will include m. Is this statement correct? Why or why not?
9.35 ▼ Acrylic bone cement is sometimes used in hip and
knee replacements to fix an artificial joint in place. The
force required to break an acrylic bone cement bond was
measured for six specimens under specified conditions,
and the resulting mean and standard deviation were
306.09 Newtons and 41.97 Newtons, respectively. Assuming that it is reasonable to assume that breaking force under these conditions has a distribution that is approximately normal, estimate the true average breaking force
for acrylic bone cement under the specified conditions.
Bold exercises answered in back
9.36 The article “The Association Between Television
Viewing and Irregular Sleep Schedules Among Children
Less Than 3 Years of Age” (Pediatrics [2005]: 851–856)
reported the accompanying 95% confidence intervals for
average TV viewing time (in hours per day) for three different age groups.
Age Group
95% Confidence Interval
Less than 12 months
12 to 23 months
24 to 35 months
(0.8, 1.0)
(1.4, 1.8)
(2.1, 2.5)
a. Suppose that the sample sizes for each of the three age
group samples were equal. Based on the given confidence
intervals, which of the age group samples had the greatest
variability in TV viewing time? Explain your choice.
b. Now suppose that the sample standard deviations for
the three age group samples were equal, but that the three
sample sizes might have been different. Which of the three
age group samples had the largest sample size? Explain
your choice.
c. The interval (.768, 1.302) is either a 90% confidence
interval or a 99% confidence interval for the mean TV
viewing time for children less than 12 months old. Is the
confidence level for this interval 90% or 99%? Explain
your choice.
9.37 Five hundred randomly selected working adults living in Calgary, Canada were asked how long, in minutes,
their typical daily commute was (Calgary Herald Traffic
Study, Ipsos, September 17, 2005). The resulting sample
mean and standard deviation of commute time were 28.5
minutes and 24.2 minutes, respectively. Construct and interpret a 90% confidence interval for the mean commute
time of working adult Calgary residents.
9.38 The article “Most Canadians Plan to Buy Treats,
Many Will Buy Pumpkins, Decorations and/or Costumes”
(Ipsos-Reid, October 24, 2005) summarized results from
a survey of 1000 randomly selected Canadian residents.
Each individual in the sample was asked how much he or
she anticipated spending on Halloween during 2005. The
resulting sample mean and standard deviation were $46.65
and $83.70 respectively.
a. Explain how it could be possible for the standard deviation of the anticipated Halloween expense to be larger
than the mean anticipated expense.
● Data set available online but not required
▼ Video solution available
9.3
b. Is it reasonable to think that the distribution of the variable anticipated Halloween expense is approximately normal? Explain why or why not.
c. Is it appropriate to use the t confidence interval to estimate the mean anticipated Halloween expense for Canadian residents? Explain why or why not.
d. If appropriate, construct and interpret a 99% confidence
interval for the mean anticipated Halloween expense for
Canadian residents.
9.39 Because of safety considerations, in May 2003 the
Federal Aviation Administration (FAA) changed its guidelines for how small commuter airlines must estimate passenger weights. Under the old rule, airlines used 180 lb
as a typical passenger weight (including carry-on luggage) in warm months and 185 lb as a typical weight in
cold months. The Alaska Journal of Commerce (May 25,
2003) reported that Frontier Airlines conducted a study
to estimate average passenger plus carry-on weights.
They found an average summer weight of 183 lb and
a winter average of 190 lb. Suppose that each of these
estimates was based on a random sample of 100 passengers and that the sample standard deviations were
20 lb for the summer weights and 23 lb for the winter
weights.
a. Construct and interpret a 95% confidence interval for
the mean summer weight (including carry-on luggage) of
Frontier Airlines passengers.
b. Construct and interpret a 95% confidence interval for
the mean winter weight (including carry-on luggage) of
Frontier Airlines passengers.
c. The new FAA recommendations are 190 lb for summer and 195 lb for winter. Comment on these recommendations in light of the confidence interval estimates from
Parts (a) and (b).
9.40 “Heinz Plays Catch-up After Under-Filling Ketchup
Containers” is the headline of an article that appeared on
CNN.com (November 30, 2000). The article stated that
Heinz had agreed to put an extra 1% of ketchup into each
ketchup container sold in California for a 1-year period.
Suppose that you want to make sure that Heinz is in fact
fulfilling its end of the agreement. You plan to take a
sample of 20-oz bottles shipped to California, measure the
amount of ketchup in each bottle, and then use the resulting data to estimate the mean amount of ketchup in each
bottle. A small pilot study showed that the amount of
ketchup in 20-oz bottles varied from 19.9 to 20.3 oz. How
many bottles should be included in the sample if you want
Bold exercises answered in back
■
Confidence Interval for a Population Mean 507
to estimate the true mean amount of ketchup to within
0.1 oz with 95% confidence?
9.41 ● ▼ Example 9.3 gave the following airborne times
for United Airlines flight 448 from Albuquerque to Denver
on 10 randomly selected days:
57
54
55
51
56
48
52
51
59
59
a. Compute and interpret a 90% confidence interval for
the mean airborne time for flight 448.
b. Give an interpretation of the 90% confidence level associated with the interval estimate in Part (a).
c. Based on your interval in Part (a), if flight 448 is
scheduled to depart at 10 A.M., what would you recommend for the published arrival time? Explain.
9.42 The authors of the paper “Short-Term Health and
Economic Benefits of Smoking Cessation: Low Birth
Weight” (Pediatrics [1999]: 1312–1320) investigated the
medical cost associated with babies born to mothers who
smoke. The paper included estimates of mean medical
cost for low-birth-weight babies for different ethnic groups.
For a sample of 654 Hispanic low-birth-weight babies, the
mean medical cost was $55,007 and the standard error
(s/ 1n) was $3011. For a sample of 13 Native American
low-birth-weight babies, the mean and standard error were
$73,418 and $29,577, respectively. Explain why the two
standard errors are so different.
9.43 ● ▼ A study of the ability of individuals to walk in a
straight line (“Can We Really Walk Straight?” American
Journal of Physical Anthropology [1992]: 19–27) reported
the following data on cadence (strides per second) for a
sample of n 20 randomly selected healthy men:
0.95 0.85 0.92 0.95 0.93 0.86 1.00 0.92 0.85 0.81
0.78 0.93 0.93 1.05 0.93 1.06 1.06 0.96 0.81 0.96
Construct and interpret a 99% confidence interval for the
population mean cadence.
9.44 ● Fat contents (in percentage) for 10 randomly selected hot dogs were given in the article “Sensory and
Mechanical Assessment of the Quality of Frankfurters”
(Journal of Texture Studies [1990]: 395–409). Use the following data to construct a 90% confidence interval for the
true mean fat percentage of hot dogs:
25.2 21.3 22.8 17.0 29.8 21.0 25.5 16.0 20.9 19.5
● Data set available online but not required
▼ Video solution available
508
Chapter 9
■
Estimation Using a Single Sample
9.45 ● Five students visiting the student health center for
a free dental examination during National Dental Hygiene
Month were asked how many months had passed since
their last visit to a dentist. Their responses were as follows:
6
17
11
22
29
Assuming that these five students can be considered a
random sample of all students participating in the free
checkup program, construct a 95% confidence interval for
the mean number of months elapsed since the last visit to
a dentist for the population of students participating in the
program.
9.46 The article “First Year Academic Success: A Prediction Combining Cognitive and Psychosocial Variables for
Caucasian and African American Students” (Journal of
College Student Development [1999]: 599–610) reported
that the sample mean and standard deviation for high
school grade point average (GPA) for students enrolled at
a large research university were 3.73 and 0.45, respectively.
Suppose that the mean and standard deviation were based
on a random sample of 900 students at the university.
a. Construct a 95% confidence interval for the mean high
school GPA for students at this university.
b. Suppose that you wanted to make a statement about the
range of GPAs for students at this university. Is it reasonable to say that 95% of the students at the university have
GPAs in the interval you computed in Part (a)? Explain.
9.47 ● ▼ The following data are the calories per half-cup
serving for 16 popular chocolate ice cream brands reviewed by Consumer Reports (July 1999):
270
190
150
190
170
160
140
170
160
150
160
110
160
180
290
170
half-cup serving of chocolate ice cream? Explain why or
why not.
9.48 The Bureau of Alcohol, Tobacco, and Firearms
(BATF) has been concerned about lead levels in California
wines. In a previous testing of wine specimens, lead levels
ranging from 50 to 700 parts per billion were recorded
(San Luis Obispo Telegram Tribune, June 11, 1991). How
many wine specimens should be tested if the BATF wishes
to estimate the true mean lead level for California wines to
within 10 parts per billion with 95% confidence?
9.49 ▼ The article “National Geographic, the Doomsday
Machine,” which appeared in the March 1976 issue of the
Journal of Irreproducible Results (yes, there really is a
journal by that name—it’s a spoof of technical journals!)
predicted dire consequences resulting from a nationwide
buildup of National Geographic magazines. The author’s
predictions are based on the observation that the number
of subscriptions for National Geographic is on the rise
and that no one ever throws away a copy of National Geographic. A key to the analysis presented in the article is
the weight of an issue of the magazine. Suppose that you
were assigned the task of estimating the average weight
of an issue of National Geographic. How many issues
should you sample to estimate the average weight to within
0.1 oz with 95% confidence? Assume that s is known to
be 1 oz.
9.50 The formula described in this section for determining sample size corresponds to a confidence level of 95%.
What would be the appropriate formula for determining
sample size when the desired confidence level is 90%?
98%?
Is it reasonable to use the t confidence interval to compute
a confidence interval for m, the true mean calories per
Bold exercises answered in back
● Data set available online but not required
▼ Video solution available
........................................................................................................................................
9.4
Interpreting and Communicating the Results of
Statistical Analyses
The purpose of most surveys and many research studies is to produce estimates of
population characteristics. One way of providing such an estimate is to construct and
report a confidence interval for the population characteristic of interest.
9.4
■
■
Interpreting and Communicating the Results of Statistical Analyses 509
Communicating the Results of a Statistical Analysis ....................................
When using sample data to estimate a population characteristic, a point estimate or a
confidence interval estimate might be used. Confidence intervals are generally preferred because a point estimate by itself does not convey any information about the
accuracy of the estimate. For this reason, whenever you report the value of a point
estimate, it is a good idea to also include an estimate of the bound on the error of
estimation.
Reporting and interpreting a confidence interval estimate requires a bit of care.
First, always report both the confidence interval and the confidence level associated
with the method used to produce the interval. Then, remember that both the confidence
interval and the confidence level should be interpreted. A good strategy is to begin
with an interpretation of the confidence interval in the context of the problem and then
to follow that with an interpretation of the confidence level. For example, if a 90%
confidence interval for p, the proportion of students at a particular university who own
a computer, is (.56, .78), we might say
interpretation of interval
explanation of “90% confidence”
interpretation of confidence level
We can be 90% confident that between
• 56% and 78% of the students at this
university own computers.
We have used a method to produce
this estimate that is successful in
µ
capturing the actual population
proportion 90% of the time.
When providing an interpretation of a confidence interval, remember that the interval is an estimate of a population characteristic and be careful not to say that the interval applies to individual values in the population or to the values of sample statistics. For example, if a 99% confidence interval for m, the mean amount of ketchup in
bottles labeled as 12 oz, is (11.94, 11.98), this does not tell us that 99% of 12-oz
ketchup bottles contain between 11.94 and 11.98 oz of ketchup. Nor does it tell us that
99% of samples of the same size would have sample means in this particular range.
The confidence interval is an estimate of the mean for all bottles in the population of
interest.
■
Interpreting the Results of a Statistical Analysis ............................................
Unfortunately, there is no customary way of reporting the estimates of population
characteristics in published sources. Possibilities include
confidence interval
estimate bound on error
estimate standard error
If the population characteristic being estimated is a population mean, then you may
also see
sample mean sample standard deviation
If the interval reported is described as a confidence interval, a confidence level
should accompany it. These intervals can be interpreted just as we have interpreted the
confidence intervals in this chapter, and the confidence level specifies the long-run er-
510
Chapter 9
■
Estimation Using a Single Sample
ror rate associated with the method used to construct the interval (e.g., a 95% confidence level specifies a 5% long-run error rate).
A form particularly common in news articles is estimate bound on error, where
the bound on error is also sometimes called the margin of error. The bound on error
reported is usually 2 times the standard deviation of the estimate. This method of reporting is a little more informal than a confidence interval and, if the sample size is
reasonably large, is roughly equivalent to reporting a 95% confidence interval. You can
interpret these intervals as you would a confidence interval with approximate confidence level of 95%.
You must use care in interpreting intervals reported in the form of an estimate standard error. Recall from Section 9.2 that the general form of a confidence interval is
estimate (critical value)(standard deviation of the estimate)
In journal articles, the estimated standard deviation of the estimate is usually referred
to as the standard error. The critical value in the confidence interval formula was determined by the form of the sampling distribution of the estimate and by the confidence level. Note that the reported form, estimate standard error, is equivalent to
a confidence interval with the critical value set equal to 1. For a statistic whose sampling distribution is (approximately) normal (such as the mean of a large sample or a
large-sample proportion), a critical value of 1 corresponds to an approximate confidence level of about 68%. Because a confidence level of 68% is rather low, you may
want to use the given information and the confidence interval formula to convert to an
interval with a higher confidence level.
When researchers are trying to estimate a population mean, they sometimes report
sample mean sample standard deviation. Be particularly careful here. To convert
this information into a useful interval estimate of the population mean, you must first
convert the sample standard deviation to the standard error of the sample mean (by dividing by 1n) and then use the standard error and an appropriate critical value to construct a confidence interval.
For example, suppose that a random sample of size 100 is used to estimate the
population mean. If the sample resulted in a sample mean of 500 and a sample standard deviation of 20, you might find the published results summarized in any of the
following ways:
95% confidence interval for the population mean: (496.08, 503.92)
mean bound on error: 500 4
mean standard error: 500 2
mean standard deviation: 500 20
■
What to Look For in Published Data ................................................................
Here are some questions to ask when you encounter interval estimates in research
reports.
■
■
Is the reported interval a confidence interval, mean bound on error, mean standard error, or mean standard deviation? If the reported interval is not a confidence interval, you may want to construct a confidence interval from the given
information.
What confidence level is associated with the given interval? Is the choice of confidence level reasonable? What does the confidence level say about the long-run
error rate of the method used to construct the interval?
9.4
■
■
Interpreting and Communicating the Results of Statistical Analyses 511
Is the reported interval relatively narrow or relatively wide? Has the population
characteristic been estimated precisely?
For example, the article “Use of a Cast Compared with a Functional Ankle Brace
After Operative Treatment of an Ankle Fracture” (Journal of Bone and Joint Surgery
[2003]: 205–211) compared two different methods of immobilizing an ankle after surgery to repair damage from a fracture. The article includes the following statement:
The mean duration (and standard deviation) between the operation and return to
work was 6313 days (median, sixty-three days; range, thirty three to ninetyeight days) for the cast group and 6519 days (median, sixty-two days; range,
eight to 131 days) for the brace group; the difference was not significant.
This is an example of a case where we must be careful—the reported intervals are
of the form estimate standard deviation. We can use this information to construct a
confidence interval for the mean time between surgery and return to work for each
method of immobilization. One hundred patients participated in the study, with 50
wearing a cast after surgery and 50 wearing an ankle brace (random assignment was
used to assign patients to treatment groups). Because the sample sizes are both large,
we can use the t confidence interval formula
mean 1t critical value2 a
s
b
1n
Each sample has df 50 1 49. The closest df value in Appendix Table 3 is
for df 40, and the corresponding t critical value for a 95% confidence level is 2.02.
The corresponding intervals are
13
b 63 3.71 159.29, 66.712
150
19
b 65 5.43 159.57, 70.432
Brace: 65 2.02 a
150
Cast: 63 2.02 a
The chosen confidence level of 95% implies that the method used to construct each
of the intervals has a 5% long-run error rate. Assuming that it is reasonable to view
these samples as representative of the patient population, we can interpret these intervals as follows: We can be 95% confident that the mean return-to-work time for those
treated with a cast is between 59.29 and 66.71 days, and we can be 95% confident that
the mean return-to-work time for those treated with an ankle brace is between 59.57
and 70.43 days. These intervals are relatively wide, indicating that the values of the
treatment means have not been estimated as precisely as we might like. This is not surprising, given the sample sizes and the variability in each sample. Note that the two
intervals overlap. This supports the statement that the difference between the two immobilization methods was not significant. Formal methods for directly comparing two
groups, covered in Chapter 11, could be used to further investigate this issue.
■
A Word to the Wise: Cautions and Limitations ..............................................
When working with point and confidence interval estimates, here are a few things you
need to keep in mind;
1. In order for an estimate to be useful, we must know something about accuracy. You
should beware of point estimates that are not accompanied by a bound on error or
some other measure of accuracy.
512
Chapter 9
■
Estimation Using a Single Sample
2. A confidence interval estimate that is wide indicates that we don’t have very precise information about the population characteristics being estimated. Don’t be
fooled by a high confidence level if the resulting interval is wide. High confidence,
while desirable, is not the same thing as saying we have precise information about
the value of a population characteristic.
The width of a confidence interval is affected by the confidence level, the
sample size, and the standard deviation of the statistic used (e.g. p or x ) as the basis for constructing the interval. The best strategy for decreasing the width of a confidence interval is to take a larger sample. It is far better to think about this before
collecting data and to use the required sample size formulas to determine a sample
size that will result in a confidence interval estimate that is narrow enough to provide useful information.
3. The accuracy of estimates depends on the sample size, not the population size.
This may be counter to intuition, but as long as the sample size is small relative
to the population size (n less than 10% of the population size), the bound on error
for estimating a population proportion with 95% confidence is approximately
p11 p 2
2
and for estimating a population mean with 95% confidence is apn
B
s
proximately 2
.
1n
Note that each of these involves the sample size n, and both bounds decrease
as the sample size increases. Neither approximate bound on error depends on the
population size.
The size of the population does need to be considered if sampling is without
replacement and the sample size is more than 10% of the population size. In this
Nn
case, a finite population correction factor
is used to adjust the bound
BN 1
on error (the given bound is multiplied by the correction factor). Since this correction factor is always less than 1, the adjusted bound on error is smaller.
4. Assumptions and “plausibility” conditions are important. The confidence interval
procedures of this chapter require certain assumptions. If these assumptions are met,
the confidence intervals provide us with a method for using sample data to estimate
population characteristics with confidence. When the assumptions associated with
a confidence interval procedure are in fact true, the confidence level specifies a correct success rate for the method. However, assumptions (such as the assumption of
a normal population distribution) are rarely exactly met in practice. Fortunately, in
most cases, as long as the assumptions are approximately met, the confidence interval procedures still work well.
In general we can only determine if assumptions are “plausible” or approximately met, and that we are in the situation where we expect the inferential procedure to work reasonably well. This is usually confirmed by knowledge of the
data collection process and by using the sample data to check certain “plausibility
conditions”.
The formal assumptions for the z confidence interval for a population proportion are
1. The sample is a random sample from the population of interest.
2. The sample size is large enough for the sampling distribution of p to be approximately normal.
3. Sampling is without replacement.
9.4
■
Interpreting and Communicating the Results of Statistical Analyses 513
Whether the random sample assumption is plausible will depend on how the sample
was selected and the intended population. Plausibility conditions for the other two
assumptions are the following:
np 10 and n(1 p) 10 (so the sampling distribution of p is approximately normal), and
n is less than 10% of the population size (so sampling with replacement
approximates sampling without replacement).
The formal assumptions for the t confidence interval for a population mean are
1. The sample is a random sample from the population of interest.
2. The population distribution is normal, so that the distribution of t xm
s/ 1n
has a t distribution.
The plausibility of the random sample assumption, as was the case for proportions,
will depend on how the sample was selected and the population of interest. The
plausibility conditions for the normal population distribution assumption are the
following:
A normal probability plot of the data is reasonably straight (indicating that
the population distribution is approximately normal), or
The data distribution is approximately symmetric and there are no outliers.
This may be confirmed by looking at a dotplot, boxplot, stem-and-leaf
display, or histogram of the data. (This would indicate that the population is approximately normal.)
Alternatively, if n is large (n 30), the sampling distribution of x will be approximately normal even for nonnormal population distributions. This implies that use
of the t interval is appropriate even if population normality is not plausible.
In the end, you must decide that the assumptions are met or that they are “plausible” and that the inferential method used will provide reasonable results. This is
also true for the inferential methods introduced in the chapters that follow.
5. Watch out for the “” when reading published reports. Don’t fall into the trap of
thinking confidence interval every time you see a in an expression. As was discussed earlier in this section, published reports are not consistent, and in addition
to confidence intervals, it is common to see estimate standard error and estimate sample standard deviation reported.
514
Chapter 9
■
A c t i v i t y 9.1
Estimation Using a Single Sample
Getting a Feel for Confidence Level
Explore this applet activity
Technology Activity (Applet): Open the applet (available
in ThomsonNOW at www.thomsonedu.com/login) called
ConfidenceIntervals. You should see a screen like the one
shown.
Getting Started: If the “Method” box does not say
“Means,” use the drop-down menu to select Means. In
the box just below, select “t” from the drop-down menu.
This applet will select a random sample from a specified
normal population distribution and then use the sample
to construct a confidence interval for the population mean.
The interval is then plotted on the display at the right,
and we can see if the resulting interval contains the actual
value of the population mean.
For purposes of this activity, we will sample from a
normal population with mean 100 and standard deviation
5. We will begin with a sample size of n 10. In the applet window, set m 100, s 5 and n 10. Leave the
conf-level box set at 95%. Click the “Recalculate” button
to rescale the picture on the right. Now click on the sample
button. You should see a confidence interval appear on
the display on the right hand side. If the interval contains
the actual mean of 100, the interval is drawn in green;
if 100 is not in the confidence interval, the interval is
shown in red. Your screen should look something like
the following.
Part 1: Click on the “Sample” button several more times,
and notice how the confidence interval estimate changes
from sample to sample. Also notice that at the bottom
of the left-hand side of the display, the applet is keeping
track of the proportion of all the intervals calculated so far
that include the actual value of m. If we were to construct
a large number of intervals, this proportion should closely
approximate the capture rate for the confidence interval
method.
To look at more than 1 interval at a time, change the
“Intervals” box from 1 to 100, and then click the sample
button. You should see a screen similar to the one at the
top of page 515, with 100 intervals is the display on the
right-hand side. Again, intervals containing 100 (the value
of m in this case) will be green and those that do not contain 100 will be red. Also note that the capture proportion
on the left-hand side has also been updated to reflect what
happened with the 100 newly generated intervals.
Activity 9.2
■
An Alternative Confidence Interval for a Population Proportion 515
Experiment with three other confidence levels of your
choice, and then answer the following question:
b. In general, is the proportion of computed t confidence
intervals that contain m 100 close to the stated confidence level?
Continue generating intervals until you have seen
at least 1000 intervals, and then answer the following
question:
a. How does the proportion of intervals constructed that
contain m 100 compare to the stated confidence level of
95%? On how many intervals was your proportion based?
(Note—if you followed the instructions, this should be at
least 1000.)
A c t i v i t y 9.2
An Alternative Confidence Interval for a
Population Proportion
Technology Activity (Applet): This activity presumes that
you have already worked through Activity 9.1.
Background: In Section 9.2, it was suggested that a
confidence interval of the form
P mod 1z critical value 2
P mod 11 P mod 2
n
B
successes 2
is an alternative to the usual
n4
large-sample z confidence interval. This alternative interval is preferred by many statisticians because, in repeated
sampling, the proportion of intervals constructed that include the actual value of the population proportion, p,
tends to be closer to the stated confidence level. In this
activity, we will explore how the “capture rates” for the
two different interval estimation methods compare.
Open the applet (available in ThomsonNOW at
www.thomsonedu.com/login) called ConfidenceIntervals.
You should see a screen like the one shown.
where Pmod Part 2: When the population is normal but s is unknown,
we construct a confidence interval for a population mean
using a t critical value rather than a z critical value. How
important is this distinction?
Let’s investigate. Use the drop-down menu to change
the box just below the method box that’s says “Means”
from “t” to “z with s.” The applet will now construct intervals using the sample standard deviation, but will use a z
critical value rather than the t critical value.
Use the applet to construct at least 1000 95% intervals, and then answer the following question:
c. Comment on how the proportion of the computed intervals that include the actual value of the population mean
compares to the stated confidence level of 95%. Is this
surprising? Explain why or why not.
Now experiment with some different samples sizes.
What happens when n 20? n 50? n 100? Use what
you have learned to write a paragraph explaining what
these simulations tell you about the advisability of using a
z critical value in the construction of a confidence interval
for m when s is unknown.
Explore this applet activity
516
Chapter 9
■
Estimation Using a Single Sample
Select “Proportion” from the Method box drop-down
menu, and then select “Large Sample z” from the dropdown menu of the second box. We will consider sampling
from a population with p .3 using a sample size of
40. In the applet window, enter p .3 and n 40.
Note that n 40 is large enough to satisfy np 10
and n(1 p) 10.
Set the “Intervals” box to 100, and then use the applet
to construct a large number (at least 1000) of 95% confidence intervals.
1. How does the proportion of intervals constructed that
include p .3, the population proportion, compare to
95%? Does this surprise you? Explain.
Now use the drop-down menu to change “Large
Sample z” to “Modified.” Now the applet will construct
A c t i v i t y 9.3
Verifying Signatures on a Recall Petition
Background: In 2003, petitions were submitted to the
California Secretary of State calling for the recall of Governor Gray Davis. Each of California’s 58 counties then
had to report the number of valid signatures on the petitions from that county so that the State could determine
whether there were enough valid signatures to certify the
recall and set a date for the recall election. The following paragraph appeared in the San Luis Obispo Tribune
(July 23, 2003):
In the campaign to recall Gov. Gray Davis, the secretary of state is reporting 16,000 verified signatures
from San Luis Obispo County. In all, the County
Clerk’s Office received 18,866 signatures on recall
petitions and was instructed by the state to check a
A c t i v i t y 9.4
the alternative confidence interval that is based on Pmod.
Use the applet to construct a large number (at least 1000)
of 95% confidence intervals.
2. How does the proportion of intervals constructed that
include p .3, the population proportion, compare to
95%? Is this proportion closer to 95% than was the case
for the large-sample z interval?
3. Experiment with different combinations of values of
sample size and population proportion p. Can you find a
combination for which the large sample z interval has a
capture rate that is close to 95%? Can you find a combination for which it has a capture rate that is even farther
from 95% than it was for n 40 and p .3? How does
the modified interval perform in each of these cases?
random sample of 567. Out of those, 84.48% were
good. The verification process includes checking
whether the signer is a registered voter and whether
the address and signature on the recall petition match
the voter registration.
1. Use the data from the random sample of 567 San Luis
Obispo County signatures to construct a 95% confidence
interval for the proportion of petition signatures that are
valid.
2. How do you think that the reported figure of 16,000 verified signature for San Luis Obispo County was obtained?
3. Based on your confidence interval from Step 1, explain
why you think that the reported figure of 16,000 verified
signatures is or is not reasonable.
A Meaningful Paragraph
Write a meaningful paragraph that includes the following
six terms: sample, population, confidence level, estimate, mean, margin of error.
A “meaningful paragraph” is a coherent piece writing
in an appropriate context that uses all of the listed words.
The paragraph should show that you understand the mean-
ing of the terms and their relationship to one another. A
sequence of sentences that just define the terms is not a
meaningful paragraph. When choosing a context, think
carefully about the terms you need to use. Choosing a
good context will make writing a meaningful paragraph
easier.
■
Summary of Key Concepts and Formulas 517
Summary of Key Concepts and Formulas
Term or Formula
Comment
Point estimate
A single number, based on sample data, that represents a
plausible value of a population characteristic.
Unbiased statistic
A statistic that has a sampling distribution with a mean
equal to the value of the population characteristic to be
estimated.
Confidence interval
An interval that is computed from sample data and
provides a range of plausible values for a population
characteristic.
Confidence level
A number that provides information on how much “confidence” we can have in the method used to construct a
confidence interval estimate. The confidence level specifies the percentage of all possible samples that will produce an interval containing the true value of the population characteristic.
p 1z critical value2
n p11 p 2 a
B
p11 p2
n
1.96 2
b
B
x 1z critical value2
s
1n
x 1t critical value2
s
1n
n a
1.96s 2
b
B
A formula used to construct a confidence interval for p
when the sample size is large.
A formula used to compute the sample size necessary
for estimating p to within an amount B with 95% confidence. (For other confidence levels, replace 1.96 with
an appropriate z critical value.)
A formula used to construct a confidence interval for m
when s is known and either the sample size is large or
the population distribution is normal.
A formula used to construct a confidence interval for m
when s is unknown and either the sample size is large or
the population distribution is normal.
A formula used to compute the sample size necessary
for estimating m to within an amount B with 95% confidence. (For other confidence levels, replace 1.96 with
an appropriate z critical value.)
518
Chapter 9
■
Estimation Using a Single Sample
Chapter Review Exercises 9.51–9.73
Know exactly what to study! Take a pre-test and receive your Personalized Learning Plan.
9.51 Despite protests from civil libertarians and gay rights
activists, many people favor mandatory AIDS testing of
certain at-risk groups, and some people even believe that
all citizens should be tested. What proportion of the adults
in the United States favor mandatory testing for all citizens? To assess public opinion on this issue, researchers
conducted a survey of 1014 randomly selected adult U.S.
citizens (“Large Majorities Continue to Back AIDS Testing,” Gallup Poll Monthly [1991]: 25–28). The article reported that 466 of the 1014 people surveyed believed that
all citizens should be tested. Use this information to estimate p, the true proportion of all U.S. adults who favor
AIDS testing of all citizens.
9.54 Seventy-seven students at the University of Virginia
were asked to keep a diary of a conversation with their
mothers, recording any lies they told during these conversations (San Luis Obispo Telegram-Tribune, August 16,
1995). It was reported that the mean number of lies per
conversation was 0.5. Suppose that the standard deviation
(which was not reported) was 0.4.
a. Suppose that this group of 77 is a random sample from
the population of students at this university. Construct a
95% confidence interval for the mean number of lies per
conversation for this population.
b. The interval in Part (a) does not include 0. Does this
imply that all students lie to their mothers? Explain.
9.52 The article “Consumers Show Increased Liking for
Diesel Autos” (USA Today, January 29, 2003) reported
that 27% of U.S. consumers would opt for a diesel car if it
ran as cleanly and performed as well as a car with a gas
engine. Suppose that you suspect that the proportion might
be different in your area and that you want to conduct a
survey to estimate this proportion for the adult residents
of your city. What is the required sample size if you want
to estimate this proportion to within .05 with 95% confidence? Compute the required sample size first using .27 as
a preliminary estimate of p and then using the conservative value of .5. How do the two sample sizes compare?
What sample size would you recommend for this study?
9.55 The article “Selected Characteristics of High-Risk
Students and Their Enrollment Persistence” (Journal of
College Student Development [1994]: 54–60) examined
factors that affect whether students stay in college. The following summary statistics are based on data from a sample
of 44 students who did not return to college after the first
quarter (the nonpersisters) and a sample of 257 students
who did return (the persisters):
9.53 In the article “Fluoridation Brushed Off by Utah”
(Associated Press, August 24, 1998), it was reported that
a small but vocal minority in Utah has been successful in
keeping fluoride out of Utah water supplies despite evidence that fluoridation reduces tooth decay and despite the
fact that a clear majority of Utah residents favor fluoridation. To support this statement, the article included the result of a survey of Utah residents that found 65% to be in
favor of fluoridation. Suppose that this result was based
on a random sample of 150 Utah residents. Construct and
interpret a 90% confidence interval for p, the true proportion of Utah residents who favor fluoridation. Is this interval consistent with the statement that fluoridation is favored by a clear majority of residents?
Bold exercises answered in back
Number of Hours Worked per Week During the
First Quarter
Nonpersisters
Persisters
Mean
Standard
Deviation
25.62
18.10
14.41
15.31
a. Consider the 44 nonpersisters as a random sample from
the population of all nonpersisters at the university where
the data were collected. Compute a 98% confidence interval for the mean number of hours worked per week for
nonpersisters.
b. Consider the 257 persisters as a random sample from
the population of all persisters at the university where
the data were collected. Compute a 98% confidence interval for the mean number of hours worked per week for
persisters.
● Data set available online but not required
▼ Video solution available
■
c. The 98% confidence interval for persisters is narrower
than the corresponding interval for nonpersisters, even
though the standard deviation for persisters is larger than
that for nonpersisters. Explain why this happened.
d. Based on the interval in Part (a), do you think that the
mean number of hours worked per week for nonpersisters
is greater than 20? Explain.
9.56 An Associated Press article on potential violent behavior reported the results of a survey of 750 workers who
were employed full time (San Luis Obispo Tribune, September 7, 1999). Of those surveyed, 125 indicated that
they were so angered by a coworker during the past year
that they felt like hitting the coworker (but didn’t). Assuming that it is reasonable to regard this sample of 750 as a
random sample from the population of full-time workers,
use this information to construct and interpret a 90% confidence interval estimate of p, the true proportion of fulltime workers so angered in the last year that they wanted
to hit a colleague.
9.57 The 1991 publication of the book Final Exit, which
includes chapters on doctor-assisted suicide, caused a
great deal of controversy in the medical community. The
Society for the Right to Die and the American Medical
Association quoted very different figures regarding the
proportion of primary-care physicians who have participated in some form of doctor-assisted suicide for terminally ill patients (USA Today, July 1991). Suppose that a
survey of physicians is to be designed to estimate this proportion to within .05 with 95% confidence. How many
primary-care physicians should be included in a random
sample?
9.58 Retailers report that the use of cents-off coupons is
increasing. The Scripps Howard News Service (July 9,
1991) reported the proportion of all households that use
coupons as .77. Suppose that this estimate was based on
a random sample of 800 households (i.e., n 800 and
p .77). Construct a 95% confidence interval for p, the
true proportion of all households that use coupons.
9.59 A manufacturer of small appliances purchases plastic handles for coffeepots from an outside vendor. If a
handle is cracked, it is considered defective and must be
discarded. A large shipment of plastic handles is received.
Bold exercises answered in back
Chapter Review Exercises 519
The proportion of defective handles p is of interest. How
many handles from the shipment should be inspected to
estimate p to within 0.1 with 95% confidence?
9.60 An article in the Chicago Tribune (August 29, 1999)
reported that in a poll of residents of the Chicago suburbs,
43% felt that their financial situation had improved during the past year. The following statement is from the article: “The findings of this Tribune poll are based on interviews with 930 randomly selected suburban residents.
The sample included suburban Cook County plus DuPage,
Kane, Lake, McHenry, and Will Counties. In a sample of
this size, one can say with 95% certainty that results will
differ by no more than 3 percent from results obtained if
all residents had been included in the poll.”
Comment on this statement. Give a statistical argument to justify the claim that the estimate of 43% is within
3% of the true proportion of residents who feel that their
financial situation has improved.
9.61 The McClatchy News Service (San Luis Obispo
Telegram-Tribune, June 13, 1991) reported on a study of
violence on television during primetime hours. The following table summarizes the information reported for four
networks:
Network
Mean Number of
Violent Acts per Hour
ABC
CBS
FOX
NBC
15.6
11.9
11.7
11.0
Suppose that each of these sample means was computed
on the basis of viewing n 50 randomly selected primetime hours and that the population standard deviation for
each of the four networks is known to be s 5.
a. Compute a 95% confidence interval for the true mean
number of violent acts per prime-time hour for ABC.
b. Compute 95% confidence intervals for the mean number of violent acts per prime-time hour for each of the
other three networks.
c. The National Coalition on Television Violence claims
that shows on ABC are more violent than those on the
other networks. Based on the confidence intervals from
● Data set available online but not required
▼ Video solution available
520
Chapter 9
■
Estimation Using a Single Sample
Parts (a) and (b), do you agree with this conclusion?
Explain.
9.62 The Chronicle of Higher Education (January 13,
1993) reported that 72.1% of those responding to a national survey of college freshmen were attending the college of their first choice. Suppose that n 500 students
responded to the survey (the actual sample size was much
larger).
a. Using the sample size n 500, calculate a 99% confidence interval for the proportion of college students who
are attending their first choice of college.
b. Compute and interpret a 95% confidence interval for
the proportion of students who are not attending their first
choice of college.
c. The actual sample size for this survey was much larger
than 500. Would a confidence interval based on the actual
sample size have been narrower or wider than the one
computed in Part (a)?
9.63 Increases in worker injuries and disability claims
have prompted renewed interest in workplace design and
regulation. As one particular aspect of this, employees required to do regular lifting should not have to handle unsafe loads. The article “Anthropometric, Muscle Strength,
and Spinal Mobility Characteristics as Predictors of the
Rating of Acceptable Loads in Parcel Sorting” (Ergonomics [1992]: 1033–1044) reported on a study involving a random sample of n 18 male postal workers. The
sample mean rating of acceptable load attained with a
work-simulating test was found to be x 9.7 kg. and the
sample standard deviation was s 4.3 kg. Suppose that in
the population of all male postal workers, the distribution
of rating of acceptable load can be modeled approximately
using a normal distribution with mean value m. Construct
and interpret a 95% confidence interval for m.
9.64 The Gallup Organization conducted a telephone survey on attitudes toward AIDS (Gallup Monthly, 1991). A
total of 1014 individuals were contacted. Each individual
was asked whether they agreed with the following statement: “Landlords should have the right to evict a tenant
from an apartment because that person has AIDS.” One
hundred one individuals in the sample agreed with this
statement. Use these data to construct a 90% confidence
interval for the proportion who are in agreement with this
statement. Give an interpretation of your interval.
Bold exercises answered in back
9.65 A manufacturer of college textbooks is interested
in estimating the strength of the bindings produced by a
particular binding machine. Strength can be measured by
recording the force required to pull the pages from the
binding. If this force is measured in pounds, how many
books should be tested to estimate with 95% confidence to
within 0.1 lb, the average force required to break the binding? Assume that s is known to be 0.8 lb.
9.66 Recent high-profile legal cases have many people
reevaluating the jury system. Many believe that juries in
criminal trials should be able to convict on less than a
unanimous vote. To assess support for this idea, investigators asked each individual in a random sample of Californians whether they favored allowing conviction by a
10–2 verdict in criminal cases not involving the death
penalty. The Associated Press (San Luis Obispo TelegramTribune, September 13, 1995) reported that 71% supported
the 10–2 verdict. Suppose that the sample size for this
survey was n 900. Compute and interpret a 99% confidence interval for the proportion of Californians who favor the 10–2 verdict.
9.67 The Center for Urban Transportation Research released a report stating that the average commuting distance
in the United States is 10.9 mi (USA Today, August 13,
1991). Suppose that this average is actually the mean of a
random sample of 300 commuters and that the sample
standard deviation is 6.2 mi. Estimate the true mean commuting distance using a 99% confidence interval.
9.68 In 1991, California imposed a “snack tax” (a sales
tax on snack food) in an attempt to help balance the state
budget. A proposed alternative tax was a 12¢-per-pack increase in the cigarette tax. In a poll of 602 randomly selected California registered voters, 445 responded that
they would have preferred the cigarette tax increase to the
snack tax (Reno Gazette-Journal, August 26, 1991). Estimate the true proportion of California registered voters
who preferred the cigarette tax increase; use a 95% confidence interval.
9.69 The confidence intervals presented in this chapter
give both lower and upper bounds on plausible values for
the population characteristic being estimated. In some instances, only an upper bound or only a lower bound is appropriate. Using the same reasoning that gave the large
● Data set available online but not required
▼ Video solution available
■
sample interval in Section 9.3, we can say that when n is
large, 99% of all samples have
m x 2.33
Use the data of Exercise 9.67 to obtain a 95% confidence
interval for the true standard deviation of commuting
distance.
s
1n
(because the area under the z curve to the left of 2.33 is
s
.99). Thus, x 2.33
is a 99% upper confidence bound
1n
for m. Use the data of Example 9.8 to calculate the 99%
upper confidence bound for the true wait time for bypass
patients in Ontario.
9.70 The Associated Press (December 16, 1991) reported
that in a random sample of 507 people, only 142 correctly
described the Bill of Rights as the first 10 amendments to
the U.S. Constitution. Calculate a 95% confidence interval
for the proportion of the entire population that could give
a correct description.
9.71 When n is large, the statistic s is approximately unbiased for estimating s and has approximately a normal
distribution. The standard deviation of this statistic when
s
the population distribution is normal is ss ⬇
which
12n
s
can be estimated by
. A large-sample confidence
12n
interval for the population standard deviation s is then
s 1z critical value2
s
12n
Bold exercises answered in back
Graphing Calculator Explorations 521
9.72 The interval from 2.33 to 1.75 captures an area
of .95 under the z curve. This implies that another largesample 95% confidence interval for m has lower limit
s
s
and upper limit x 1.75
. Would you
x 2.33
1n
1n
recommend using this 95% interval over the 95% interval
s
discussed in the text? Explain. (Hint: Look
x 1.96
1n
at the width of each interval.)
9.73 The eating habits of 12 bats were examined in the
article “Foraging Behavior of the Indian False Vampire
Bat” (Biotropica [1991]: 63–67). These bats consume insects and frogs. For these 12 bats, the mean time to consume a frog was x 21.9 min. Suppose that the standard
deviation was s 7.7 min. Construct and interpret a 90%
confidence interval for the mean suppertime of a vampire bat whose meal consists of a frog. What assumptions
must be reasonable for the one-sample t interval to be
appropriate?
Do you need a live tutor
for homework problems?
● Data set available online but not required
Are you ready? Take your
exam-prep post-test now.
▼ Video solution available
Graphing Calculator Explorations
E x p l o r a t i o n 9.1 The Confidence Interval for a
Population Proportion
Because confidence intervals are widely used, it will come as no surprise to you that
your calculator may have a built-in capability to determine a confidence interval for a
population proportion. Once again you will need to navigate your calculator’s menu
system, looking for key words such as INTR for “interval,” or possibly TESTS. (Confidence intervals are frequently associated with “hypothesis tests,” a topic to come
later in Chapter 10.) Once you find the right menu, look for words like “1” and “prop”
and “z”—together these key words should indicate a one-sample z confidence interval
for a proportion.
522
Chapter 9
■
Estimation Using a Single Sample
1-Prop Zinterval
C-Level :
x
:
n
:
Execute
1-PropZInt
x :
n :
C-Level:
Calculate
Common
screen presentations for
confidence interval data
entry.
Figure 9.10
Once you select the correct choice, you will be presented with a screen for providing the information needed
1-Prop Zinterval
to calculate a confidence interval: the number of sucLeft 0.42889
cesses, the sample size, and the confidence level. Two
Right0.49024
representative screens are given in Figure 9.10.
p̂ 0.46
From Example 9.4, supply the following informan1014
tion: x 466, n 1014, and the C-Level .95. Move
your cursor down to Execute or Calculate, and press the
Enter, Execute, or Calculate button, depending on your
1-PropZInt
calculator. The confidence interval should appear imme(.42889, .49024)
diately. Again, two representative screens are given in
p̂.459566075
Figure 9.11.
n1014
Notice that the formatting in these screens is slightly
different. Which of these is the “right” format? Probably
F i g u r e 9 . 1 1 Calcuneither one! Recall a previous graphing calculator explolator screens for conration in Chapter 3, where we discussed the differences
fidence intervals.
between a calculator’s presentation of information and the
appropriate way to communicate this information. Check
with your instructor to ascertain her or his preferences
about what format information is required. You will almost certainly want to report
more information than what is presented on these calculator screens. Specifically, you
will notice that the calculator does not verify the appropriateness of the large-sample
assumption! Some confidence intervals may stray outside the interval from 0 to 1,
which makes sense to the calculator, but not to the statistician. Thus, you may have to
modify the interval returned by your calculator. Also, don’t forget that your calculator
does not know the context of the problem; you will have to provide that contextual information irrespective of the form of the calculator presentation. Remember: The calculator only does the calculation; you must do the thinking.
E x p l o r a t i o n 9.2 A Confidence Interval for a
Population Mean
Finding the confidence interval for a single mean on your calculator will require you to
navigate the menu system much as you did to find the confidence interval for a proportion. The confidence interval for the mean will have some added challenges, however:
1. You will need to decide whether to base the interval on the z or t distribution, and
2. You may use previous calculations of the sample mean and standard deviation, or
the calculator will evaluate these statistics from data contained in a List.
We will use the data from Exercise 9.43 to construct a 99% confidence interval.
We have entered the data in List1 and are ready to proceed. Since the population standard deviation is not known, we will use the t confidence interval. For the t confidence
interval, the normality of the population becomes an issue. The original data are at
hand, and we can assess the plausibility of a normal population via a normal probability plot. Figure 9.12 shows the normal probability plot.
After verifying the plausibility of the normality of the population, we can now
construct the confidence interval. In this case, we have a choice of entering the sample
■
Graphing Calculator Explorations 523
(a)
(a)
(b)
(b)
(a) Edit
window for normal probability plot; (b) normal probability plot.
Figure 9.12
(a) Confidence interval input screen
for raw data; (b) confidence
interval input screen for
summary statistics.
Figure 9.13
calculations or letting the calculator evaluate the sample mean and standard deviation.
Based on that choice, we see one of the screens shown in Figure 9.13.
Figure 9.14 shows one calculator’s version of the confidence interval. Once again,
we caution you that the calculator only calculates—you must still do the thinking and
present the solution in the context of the particular problem at hand.
Calculated
confidence interval.
Figure 9.14
This page intentionally left blank

Similar documents

×

Report this document