Chapter 9

371
CHAPTER 9 ANSWERS
Exercises 9.1
9.1 A hypothesis is a statement that something is true.
9.2 The decision criterion specifies whether or not the null hypothesis should
be rejected in favor of the alternative hypothesis.
9.3 (a) The population mean is equal to some fixed amount; i.e., =
0
.
(b) The population mean is greater than 0; i.e., Ha: >
0
.
The population mean is less than
0
; i.e., Ha: <
0
.
The population mean is unequal to
0
; i.e., H
a
: =
0
.
9.4 (a) Ha: =
0
; two-tailed
(b) Ha: <
0
; left-tailed
(c) H
a
: >
0
; right-tailed
9.5 Let denote the mean cadmium level in Boletus pinicola mushrooms.
(a) H0: = 0.5 ppm (b) Ha: > 0.5 ppm (c) right-tailed test
9.6 Let denote the mean retail price of agriculture books.
(a) H0: = $66.52 (b) Ha: = $66.52 (c) two-tailed
9.7 Let denote the mean daily intake of iron by adult females under age 51.
(a) H0: = 18 mg/day (b) Ha: < 18 mg/day (c) left-tailed
9.8 Let denote the mean age of early-onset dementia.
(a) H
0
: = 55 years old (b) H
a
: < 55 years old
(c) left-tailed
9.9 Let denote the mean length of imprisonment for motor-vehicle theft
offenders in Australia.
(a) H
0
: = 16.7 months (b) H
a
: = 26.7 months
(c) two-tailed
9.10 Let denote the mean post-work heart rate of casting workers.
(a) H0: = 72 beats/min (b) Ha: > 72 beats/min
(c) right-tailed
9.11 Let denote the mean body temperature of healthy humans.
(a) H0: = 98.6F (b) Ha: = 98.6F (c) two-tailed
9.12 Let denote the mean annual salary of classroom teachers in Hawaii.
(a) H
0
: = $45.9 thousand (b) H
a
: < $45.9 thousand
(c) left-tailed
9.13 Let denote the mean local monthly bill for cell phone users in the U.S.
(a) H0: = $47.37 (b) Ha: > $47.37 (c) right-tailed
9.14 Let denote the mean annual energy consumed per U.S. household.
372 Chapter 9, Hypothesis Tests for One Population Mean
(a) H
0
: = 92.2 million BTU (mean western household energy consumption is
the same as all American households)
H
a: = 92.2 million BTU (mean western household energy consumption
differs from that of all American households)
(b) If the sample mean energy consumption x
_
differs by too much from 92.2
million BTU, then we should be inclined to reject H
0 and conclude that
Ha is true. From the data, we compute x
_
== 79.65 million BTU. The
question is whether the difference of 12.55 million BTU between the
sample mean of 79.65 million BTU and the hypothesized population mean
of 92.2 million BTU can be attributed to sampling error or whether the
difference is large enough to indicate that the population mean is not
92.2 million BTU.
(c) The sampling distribution of x
_
will be a normal distribution.
(d) It is quite unlikely that the sample mean x
_
will be more than two
standard deviations away from the population mean . If x
_
is more
than two standard deviations away from , then reject H0 and conclude
that Ha is true. Otherwise, do not reject H0.
(e) We have o = 15, n = 20, x
_
= 79.65, and = 92.2 under H0 true. Thus,
(79.65 92.2)/(15/ 20) 3.74 z = = . Since x
_
is more than two standard
deviations away from 92.2 million BTU, we reject H0 and conclude that Ha
is true.
9.15 (a) H0: = 5.6 radios per U.S. household
Ha: = 5.6 radios per U.S. household
(b) If the sample mean number of radios per U.S. household x
_
differs by too
much from 5.6 radios, then we should be inclined to reject H0 and
conclude that Ha is true. From the data, we compute x
_
== 5.89 radios.
The question is whether the difference of 0.29 radios between the
sample mean of 5.89 and the hypothesized population mean of 5.6 can be
attributed to sampling error or whether the difference is large enough
to indicate that the population mean is not 5.6 radios.
(c) The sampling distribution of x
_
will be approximately a normal
distribution.
(d) It is quite unlikely that the sample mean x
_
will be more than two
standard deviations away from the population mean . If x
_
is more
than two standard deviations away from , then reject H
0
and conclude
that Ha is true. Otherwise, do not reject H0.
(e) We have o = 1.9, n = 45, x
_
= 5.89, and = 5.6 under H0 true. Thus,
(5.89 5.6)/(1.9/ 45) 1.02 z = = . Since x
_
is less than two standard
deviations away from 5.6 radios, we do not reject H0 and conclude that
H0 is reasonable.
9.16 (a) If the mean weight x
_
of the 50 bags of pretzels sampled is more than
one standard deviation away from 454 grams, then reject the null
hypothesis that = 454 grams and conclude that the alternative
hypothesis, which is = 454 grams, is true. Otherwise, do not reject
the null hypothesis.
Graphically, the decision criterion looks like:
Section 9.1, The Nature of Hypothesis Testing 373
Do not
Reject H
0
Reject H
0
Reject H
0
454-1
454+1

-1 0 1 z
0.1587 0.6826 0.1587
454
Reject H
0
Do not Reject H
0
Reject H
0
454-1
454+1

-1 0 1 z
(b)
The lower figure shows that, using our decision criterion, the
probability is 0.3174 (= 1 - 0.6826 = 0.1587 + 0.1587) of rejecting the
null hypothesis if it is in fact true.
(c) We have o = 7.9, n = 25, x
_
= 450, and = 454 if H0 is true. Thus,
56 . 2 ) 25 / 8 . 7 /( ) 454 450 ( = = z . The sample mean x
_
is 2.56 standard
deviations below the null hypothesis mean of 454 grams. Since the mean
weight x
_
of 25 bags of pretzels sampled is more than one standard
deviation away from 454 grams, we reject the null hypothesis that =
454 grams and conclude that the alternative hypothesis, which is =
454 grams, is true. In other words, the data provide sufficient
evidence to conclude that the packaging machine is not working
properly.
9.17 (a) If the mean weight x
_
of the 25 bags of pretzels sampled is more than
three standard deviations away from 454 grams, then reject the null
hypothesis that = 454 grams and conclude that the alternative
hypothesis, which is = 454 grams, is true. Otherwise, do not reject
Graphically, the decision criterion looks like:
-4 -3 -2 -1 0 1 2 3 4
0.0013 0.9974 0.0013
Reject Do not Reject H0 Reject
H0 H0
454 3o 454 454 + 3o Mean
454
Reject H
0
Do not Reject H
0
Reject H
0
454-3
454+3

-3 0 3 z
(b)
The lower figure shows that, using our decision criterion, the
probability is 0.0026 (= 1 - 0.9974 = 0.0013 + 0.0013) of rejecting the
null hypothesis if it is in fact true.
(c) We have o = 7.9, n = 25, x
_
= 450, and = 454 if H0 is true. Thus,
56 . 2 ) 25 / 8 . 7 /( ) 454 450 ( = = z . The sample mean x
_
is 2.56 standard
deviations below the null hypothesis mean of 454 grams. Since the mean
weight x
_
of 25 bags of pretzels sampled is less than three standard
deviations away from 454 grams, we do not reject the null hypothesis
that = 454 grams and conclude that the null hypothesis, which is =
454 grams, is reasonable. In other words, the data do not provide
sufficient evidence to conclude that the packaging machine is not
working properly.
9.18 If the null hypothesis is true, the chance of incorrectly rejecting it is
0.0456 when using the 95.44% part of the 68.26-95.44-99.74 rule.
Exercises 9.2
9.19 (a) This statement is true: If it is important not to reject a true null
hypothesis, i.e., not to make a Type I error, then the hypothesis test
should be performed at a small significance level. This can be
appreciated by considering the meaning of the significance level. The
significance level is equal to the probability of making a Type I
error. The smaller the significance level, the smaller the probability
of rejecting a true null hypothesis.
(b) This statement is true: Decreasing the significance level results in
an increase in the probability of making a Type II error. This can be
appreciated by considering the relation between Type I and Type II
error probabilities. For a fixed sample size, the smaller the Type I
error probability (which is equal to the significance level), the
Section 9.2, Terms, Errors, and Hypotheses 375
larger the Type II error probability.
9.20 (a) A test statistic is a statistic used to decide whether to reject the
null hypothesis.
(b) The rejection region is a set of values of the test statistic that lead
to rejection of the null hypothesis.
(c) The nonrejection region is a set of values of the test statistic that
do not lead to rejection of the null hypothesis.
(d) Critical values are values of the test statistic that separate the
rejection region from the nonrejection region. They are considered to
be part of the rejection region.
(e) The significance level is the probability of rejecting a true null
hypothesis.
9.21 A Type I error is made when a true null hypothesis is rejected. The
probability of making this error is denoted by . A Type II error is made
when a false null hypothesis is not rejected. We denote the probability of
a Type II error by .
9.22 (a) If the null hypothesis is rejected, we conclude that the alternative
hypothesis is true.
(b) If the null hypothesis is not rejected, we conclude that the data did
not provide sufficient evidence to conclude that the alternative
hypothesis is true. We cannot say that the null hypothesis is true,
only that it is reasonable.
9.23 (a) Rejection region: z > 1.645
(b) Nonrejection region: z < 1.645
(c) Critical value: z = 1.645
(d) Significance level: o = 0.05
(e)
(f) Right-tailed test
9.24 (a) Rejection region: z s -1.96 or z > 1.96
(b) Nonrejection region: -1.96 < z < 1.96
(c) Critical values: z = 1.96
Do not
Reject H
0
Reject H
0
0 1.645 z
0.9500 0.0500
Critical Value
Nonrejection region |Rejec tion region
Do not
Reject H
0
Reject H
0
Reject H
0
-1.96 0 1.96 z
Critical value Critical value
Rejection | Nonrejectio n | Rejection
Region | Region | Region
0.0250 0.9500 0.0250
(e)
(f) Two-tailed test
9.25 (a) Rejection region: z s -2.33
(b) Nonrejection region: z > -2.33
(c) Critical value: z = -2.33
(e)
(f) Left-tailed test
9.26 (a) Rejection region: z s -1.645
(b) Nonrejection region: z > -1.645
(c) Critical value: z = -1.645
Do not
Reject H
0
Reject H
0
-2.33 0 z
Critical value
Rejection | Nonrejecti on
Region | Region
0.01 0.99
Do not
Reject H 0 Reject H0
-1.645 0 z
Critical value
Rejection | Nonrejecti on
Region | Region
0.05 0.95
(e)
(f) Left-tailed test
9.27 (a) Rejection region: z s -1.645 or z > -1.645
(b) Nonrejection region: -1.645 < z < -1.645
(c) Critical values: z = -1.645 and z = 1.645
(e)
(f) Two-tailed test
9.28 (a) Rejection region: z > 1.28
Do not
Reject H
0
Reject H
0
Reject H
0
-1.645 0 1.645 z
Rejection | Nonrejectio n | Rejection
0.0500 0.9000 0.0500
(e)
(f) Right tailed test
9.29 (a) A Type I error would occur if, in fact, = 0.5 ppm, but the results
of the sampling lead to the conclusion that > 0.5 ppm.
(b) A Type II error would occur if, in fact, > 0.5 ppm, but the results
of the sampling fail to lead to that conclusion.
(c) A correct decision would occur if, in fact, = 0.5 ppm and the
results of the sampling do not lead to the rejection of that fact; or
if, in fact, > 0.5 ppm and the results of the sampling lead to that
conclusion.
(d) If, in fact, the mean cadmium level in Boletus pinicola mushrooms is
equal to 0.5 ppm, and we do not reject the null hypothesis that =
0.5 ppm, we made a correct decision.
(e) If, in fact, the mean cadmium level in Boletus pinicola mushrooms is
greater than to 0.5 ppm, and we do not reject the null hypothesis that
= 0.5 ppm, we made a Type II error.
9.30 (a) A Type I error would occur if, in fact, = $66.52, but the results of
the sampling lead to the conclusion that = $66.52.
(b) A Type II error would occur if, in fact, = $66.52, but the results
(c) A correct decision would occur if, in fact, = $66.52 and the results
of the sampling do not lead to the rejection of that fact; or if, in
fact, = $66.52 and the results of the sampling lead to that
conclusion.
(d) If, in fact, the mean retail price of agricultural books is equal to
$66.52, and we reject the null hypothesis that = $66.52, we made a
Type I error.
(e) If, in fact, the mean retail price of agricultural books is not $66.52,
and we reject the null hypothesis that = $66.52, we made a correct
decision.
9.31 (a) A Type I error would occur if, in fact, = 18 mg, but the results of
Do not
Reject H
0
Reject H
0
0 1.28 z
0.9000 0.1000
Critical Value
Nonrejection region | Rejection region
the sampling lead to the conclusion that < 18 mg.
(b) A Type II error would occur if, in fact, < 18 mg, but the results of
the sampling fail to lead to that conclusion.
(c) A correct decision would occur if, in fact, = 18 mg and the results
fact, < 18 mg and the results of the sampling lead to that
conclusion.
(d) If the mean iron intake equals the RDA of 18 mg, and we reject the null
hypothesis that = 18 mg, we made a Type I error.
(e) If, in fact, the mean iron intake is less than the RDA of 18 mg, and we
reject the null hypothesis that = 18 mg, we made a correct decision.
9.32 (a) A Type I error would occur if, in fact, = 55 years old, but the
results of the sampling lead to the conclusion that < 55 years old.
(b) A Type II error would occur if, in fact, < 55 years old, but the
results of the sampling fail to lead to that conclusion.
(c) A correct decision would occur if, in fact, = 55 years old and the
if, in fact, < 55 years old and the results of the sampling lead to
that conclusion.
(d) If the mean age of all people with early-onset dementia is 55 years,
and we do not reject the null hypothesis that = 55 years old, we
made a correct decision.
(e) If, in fact, the mean age of all people with early-onset dementia is
less than 55 years, and we do not reject the null hypothesis that =
55 years old, we made a Type II error.
9.33 (a) A Type I error would occur if, in fact, = 16.7 months, but the
results of the sampling lead to the conclusion that = 16.7 months.
(b) A Type II error would occur if, in fact, = 16.7 months, but the
(c) A correct decision would occur if, in fact, = 16.7 months and the
if, in fact, = 16.7 months and the results of the sampling lead to
that conclusion.
(d) If, in fact, the mean length of imprisonment equals 16.7 months, and we
do not reject the null hypothesis that = 16.7 months, we made a
correct decision.
(e) If, in fact, the mean length of imprisonment does not equal 16.7
months, and we do not reject the null hypothesis that = 16.7 months,
we made a Type II error.
9.34 (a) A Type I error would occur if, in fact, = 72 bpm, but the results of
the sampling lead to the conclusion that > 72 bpm.
(b) A Type II error would occur if, in fact, > 72 bpm, but the results
(c) A correct decision would occur if, in fact, = 72 bpm and the results
fact, > 72 bpm and the results of the sampling lead to that
conclusion.
(d) If, in fact, the mean post-work heart rate of casting workers equaled
the normal resting heart rate of 72 bpm, and we rejected the null
hypothesis that = 72 bpm, we have made a Type I error.
(e) If, in fact, the mean post-work heart rate of casting workers did
exceed the normal resting heart rate of 72 bpm, and we rejected the
null hypothesis that = 72 bpm, we have made a correct decision.
9.35 (a) A Type I error would occur if, in fact, = 98.6 F, but the results
of the sampling lead to the conclusion that = 98.6 F.
(b) A Type II error would occur if, in fact, = 98.6 F, but the results
(c) A correct decision would occur if, in fact, = 98.6 F and the
if, in fact, = 98.6 F and the results of the sampling lead to that
conclusion.
(d) If the mean temperature of all healthy humans equals 98.6 F, and we
reject the null hypothesis that = 98.6 F, we made a Type I error.
(e) If, in fact, the temperature of all healthy humans is not equal to
98.6 F, and we reject the null hypothesis that = 98.6 F, we made a
correct decision.
9.36 (a) A Type I error would occur if, in fact, = $45.9 thousand, but the
results of the sampling lead to the conclusion that < $45.9
thousand.
(b) A Type II error would occur if, in fact, < $45.9 thousand, but the
(c) A correct decision would occur if, in fact, = $45.9 thousand and the
if, in fact, < $45.9 thousand and the results of the sampling lead
to that conclusion.
(d) If the mean annual salary of classroom teachers in Hawaii equals the
national mean of $45.9 thousand, and we do not reject the null
hypothesis that = $45.9 thousand, we made a correct decision.
(e) If, in fact, the mean annual salary of classroom teachers in Hawaii is
less than the national mean of $45.9 thousand, and we do not reject the
null hypothesis that = $45.9 thousand, we made a Type II error.
9.37 (a) A Type I error would occur if, in fact, = $47.37, but the results of
the sampling lead to the conclusion that > $47.37.
(b) A Type II error would occur if, in fact, > $47.37, but the results
(c) A correct decision would occur if, in fact, = $47.37 and the results
fact, > $47.37 and the results of the sampling lead to that
conclusion.
(d) If the mean phone bill equals the 2001 mean of $47.37, and we do not
reject the null hypothesis that = $47.37, we made a correct
decision.
(e) If, in fact, the mean cell phone bill is greater than the 2001 mean of
$47.37, and we do not reject the null hypothesis that = $47.37, we
made a Type II error.
9.38 (a) P(Type I error) = o = 0.
(b) If o = 0, the null hypothesis will never be rejected. If the null
hypothesis is not true, it will not be rejected. Therefore, P(Type II
error) = | = 1.
9.39 (a) Exercise 9.31 is a situation in which it may be important to have a
small o probability. Concluding that females under the age of 51 are,
on the average, getting less than the RDA of 18 mg of iron could lead
to remedial action by the nations health providers which would be
expensive and unnecessary if females under 51 are, in fact, getting the
RDA of 18 mg of iron.
(b) Exercise 9.29 is a situation in which it may be important to have a
small | probability. If the mean cadmium level in the mushrooms is,
in fact, higher than the government recommended limit, eating the
mushrooms could have serious health consequences for those who eat
them. If a hypothesis test does not lead to the conclusion that the
cadmium level is too high, the population would be led to believe that
the mushrooms are safe to eat when, in fact, they are not. The
probability of this happening is | and should be kept small.
(c) Exercise 9.40 provides a situation in which it is important to have
both small o and | probability. The null hypothesis is that the
nuclear power plant is safe and the alternative is that it is not safe.
A discussion concerning the desirability of | being small is found in
that exercise. Given the growing need for power and the power crises
in California in 2001 and in the eastern U.S. in 2003, it would also be
important not to reject a power plant (of any kind) that was actually
safe. Thus we would also want o to be small in this case.
Another situation in which small o and | are desirable involves
defective products. Suppose that a manufacturing company samples items
from boxes to determine if the percentage of defective items is too
high. The null hypothesis in each case is that the percentage of
defects is, say, 3%, while the alternative hypothesis is that the
percentage is more than 3%. If the sample results in the conclusion
that the percentage of defects is too high when, in fact, it is not, a
Type I error has been committed and the manufacturer will unnecessarily
incur the expense of shutting down an assembly line to find a
nonexistent problem. If the sample results in the conclusion that the
percentage of defects is acceptable when, in fact, it is too high, a
Type II error has been committed and the manufacturer is likely to
incur a loss of business and reputation at the hands of unhappy
customers.
9.40 In this exercise, we are told that failing to reject the null hypothesis
corresponds to approving the nuclear reactor for use. This action
approving the nuclear reactor suggests that the null hypothesis must be
something like: "The nuclear reactor is safe." This further suggests that
the alternative hypothesis is something like: "The nuclear reactor is
unsafe." Putting things together, the Type II error in this situation is:
"Approving the nuclear reactor for use when, in fact, it is unsafe." This
type of error has consequences that are catastrophic. Thus, the property
that we want the Type II error probability to exhibit is that it be small.
9.41 (a) A Type I error would occur if, in fact, the defendant is innocent, but
the jury concludes that the defendant is guilty.
(b) A Type II error would occur if, in fact, the defendant is guilty, but
the jury fails to conclude that the defendant is guilty.
(c) If I were a defendant, I would want o to be small. Given that I am
innocent, I certainly want there to be a small probability of the jury
rejecting my innocence (i.e., finding me guilty).
(d) If I were a prosecutor, I would want | to be small. Given that the
defendant is guilty, I want there to be a small probability that the
jury would declare the defendant not guilty.
(e) If o = 0, then an innocent person would never be declared guilty. If
= 0, then a guilty person would always be found guilty.
9.42 (a) The probability of a Type I error is the same as the significance
level, o = 0.0456.
(b) If the mean net weight being packaged is 447 g, then the distribution
of x
_
is a normal distribution with mean 447 g and standard deviation
. 56 . 1 25 / 8 . 7 / g n = = o
(c) | is the probability of not rejecting the null hypothesis when it is
actually false. In this case, | is the probability that x
_
falls
between 450.88 g and 457.12 g when = 447 grams and o = 1.56 g.
Thus
| = < < =
< <
P x P z ( . . ) (
.
.
.
.
) 450 88 457 12
45088 447
156
457 12 447
156
= P(2.49 < z < 6.49) = 1.0000 - 0.9936 = 0.0064
(d) The probability of a Type II error is an area between the two critical
values of x
_
(subscripts indicate left and right) above (i.e., between x
_
l
= 450.88 g and x
_
r = 457.12 g) assuming that the true mean is any one of
the thirteen values of presented in this part of the exercise. As a
probability statement, this is written P(450.88< x
_
< 457.12).
Since the sample size in this exercise is large enough (i.e., n = 25),
the random variable x
_
is approximately normally distributed with mean
=
x
and standard deviation /
x
n o o = . Thus, in order to
calculate P(450.88 < <457.12), we implement the z-score formulas
z
x
n
x
n
l a r a
=
=

o
o / /
and z
,
insert the necessary elements into the right-hand side of each formula
itself, and proceed with using Table II to find the appropriate areas.
Notice that x
_
l = 450.88 and x
_
r = 457.12 and that the standard deviation
to be inserted into each formula has already been presented; i.e.,
. 56 . 1 25 / 8 . 7 / g n = = o Most importantly, the value of the population
mean to be inserted into each formula is not the value of assuming
that the null hypothesis is true; i.e., it is not
0
= 454. It is,
instead, an alternative value of , as indicated by the symbol
a
in
each of the formulas.
For this part of the exercise, we are given thirteen alternative "true
mean" values for . This translates into 26 z-scores that need to be
computed (i.e., two for each value of the "true mean"). In turn, we
calculate the area associated with each pair of z-scores and then use
this information to compute | , defined as the probability of a Type II
error.
The appropriate calculations are:
True mean z-score P(Type II error)
computation |
448
85 . 5
25 / 8 . 7
448 12 . 457
85 . 1
25 / 8 . 7
448 88 . 450
=
=
=
=
z
z
1.0000 - 0.9678 = 0.0322
449
21 . 5
25 / 8 . 7
449 12 . 457
21 . 1
25 / 8 . 7
449 88 . 450
=
=
=
=
z
z
1.0000 - 0.8869 = 0.1131
450
56 . 4
25 / 8 . 7
450 12 . 457
56 . 0
25 / 8 . 7
450 88 . 450
=
=
=
=
z
z
1.0000 - 0.7123 = 0.2877
451
92 . 3
25 / 8 . 7
451 12 . 457
08 . 0
25 / 8 . 7
451 88 . 450
=
=
=
=
z
z
1.0000- 0.4681 = 0.5319
452
28 . 3
25 / 8 . 7
452 12 . 457
72 . 0
25 / 8 . 7
452 88 . 450
=
=
=
=
z
z
0.9995 - 0.2358 = 0.7637
453
64 . 2
25 / 8 . 7
453 12 . 457
36 . 1
25 / 8 . 7
453 88 . 450
=
=
=
=
z
z
0.9959 - 0.0869 = 0.9090
True mean z-score P(Type II error)
computation |
455
36 . 1
25 / 8 . 7
455 12 . 457
64 . 2
25 / 8 . 7
455 88 . 450
=
=
=
=
z
z
0.9131 - 0.0041 = 0.9090
456
72 . 0
25 / 8 . 7
456 12 . 457
28 . 3
25 / 8 . 7
456 88 . 450
=
=
=
=
z
z
0.7642 - 0.0005 = 0.7637
457
08 . 0
25 / 8 . 7
457 12 . 457
92 . 3
25 / 8 . 7
457 88 . 450
=
=
=
=
z
z
0.5319 - 0.0000 = 0.5319
458
56 . 0
25 / 8 . 7
458 12 . 457
56 . 4
25 / 8 . 7
458 88 . 450
=
=
=
=
z
z
0.2877 - 0.0000 = 0.2877
459
21 . 1
25 / 8 . 7
459 12 . 457
21 . 5
25 / 8 . 7
459 88 . 450
=
=
=
=
z
z
0.1131 - 0.0000 = 0.1131
460
85 . 1
25 / 8 . 7
460 12 . 457
85 . 5
25 / 8 . 7
460 88 . 450
=
=
=
=
z
z
0.0322 0.0000 = 0.0322
461
49 . 2
25 / 8 . 7
461 12 . 457
49 . 6
25 / 8 . 7
461 88 . 450
=
=
=
=
z
z
0.0064 - 0.0000 = 0.0064
To summarize this part of the exercise, notice that the answer for each |
value is presented in the third column of the previous table.
(e) Consider columns 1 and 3 of the table in part (e). Also consider a
graph whose vertical axis is labeled and whose horizontal axis is
labeled . Plot the points of | in column 3 of the table versus the
respective values of in column 1 and then connect the points with a
smooth curve. This curve is presented below.
Recall that the value of , assuming that the null hypothesis is true,
is
0
= 454 g. The previous graph tells us that the farther the true
value of is from the null hypothesis value of 454 g, the smaller is
the probability of making a Type II error; i.e., the smaller is | .
All of this is reasonable. It is more likely for a false null
hypothesis to be detected--and hence | to be small--when the true
value of is far from the null hypothesis value than when it is
close.
9.43 (a) Answers will vary. This exercise can easily be done with Minitab, but
we will describe a procedure using Excel. With a blank spreadsheet,
from the Menu bar, select Tools, Data Analysis, Random Number
Generation. Enter 100 for the number of variables, and 25 for the
Number of Random Numbers. Select Normal for the Distribution and enter
454 for the Mean and 7.8 for the Standard Deviation. Click on Output
Range and enter A1. Then click OK. This will generate 100 columns of
25 random normal numbers each.
(b) At the bottom of column A in A27, enter =AVERAGE(A1:A25) and copy this
formula into B27 through DV27.
(c) Then in A28, enter the formula =(A27-454)/(7.8/sqrt(25)), and copy this
formula into B28 through DV28. The null hypothesis will be rejected
whenever the number in row 28 is less than or equal -2 or greater than
or equal to +2.
(d) Since the significance level is 0.0456, we would expect about 4 or 5 of
the 100 samples to result in the rejection of the null hypothesis.
(e) Answers will vary. Our simulation led to rejection of the null
hypothesis 7 times.
(f) If your answer to part (d) is not 4 or 5, it is most likely the result
of sampling variation. On the average, we would expect 4.56% of all
samples to lead to rejection of the null hypothesis, but in any single
set of 100 samples, the percentage may differ from that amount.
Probabilityof aType IIError
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
445 450 455 460 465
Mean
B
e
t
a
Do not
Reject H
0
Reject H
0
Reject H
0
-1.645 0 1.645 z
Rejection | Nonreject ion | Rejection
0.0500 0.9000 0.0500
Do not
Reject H
0
Reject H
0
0 1.645 z
0.9500 0.0500
Critical Value
Nonrejection region | Rejection region
Do not
Reject H0 Reject H0
-1.645 0 z
0.0500 0.9500
Critical Value
Rejection region | Nonrejection region
Exercises 9.3
9.44 Critical values: +z
0.05
= +1.645 9.45 Critical value: z
0.05
= 1.645
9.46 Critical value: -z
0.01
= -2.33 9.47 Critical value: -z
0.05
= -1.645
9.48 Critical value: z
0.01
= 2.33 9.49 Critical values: +z
0.025
= +1.96
9.50 There are two reasons for
concern. When there are outliers, the normality assumption may be
Do not
Reject H
0
Reject H
0
0 2.33 z
0.9900 0.0100
Critical Value
Nonrejection region |R ejection region
Do not
Reject H
0
Reject H
0
Reject H
0
-1.96 0 1.96 z
Rejection | Nonrejection | Rejection
0.0250 0.9500 0.0250
Do not
Reject H
0
Reject H
0
-2.33 0 z
0.0100 0.9900
Critical Value
Rejection region |No nrejection region
Section 9.3, Hyp. Tests for One Pop. Mean When is Known 387
questioned, and even for large samples, the presence of one or more outliers
can affect the results of a z-test because the sample mean can be highly
affected by outliers.
9.51 (a) The z-test in not an appropriate method for highly skewed data when the
sample size is less than 30.
(b) The z-test is appropriate for large samples with no outliers even if
the data are mildly skewed.
9.52 (a) The z-test can be used for small samples that are close to being
normally distributed, so it is appropriate in this case.
(b) The z-test is not appropriate for these data which have an outlier and
a moderate sample size.
9.53 Reject H
0
if z < -1.645; (20 22)/(4/ 32) 2.83 z = = ; therefore, reject H
0
and
conclude that < 22.
9.54 Reject H0 if z < -1.645; (21 22)/(4/ 32) 1.41 z = = ; therefore, do not reject
H0. The data do not provide sufficient evidence to support Ha: < 22.
9.55 Reject H
0
if z > 1.645; (24 22)/(4/ 15) 1.94 z = = ; therefore, reject H
0
and
conclude that > 22.
9.56 Reject H0 if z > 1.645; (23 22)/(4/ 15) 0.97 z = = ; therefore, do not reject
H0. The data do not provide sufficient evidence to support Ha: > 22.
9.57 Reject H
0
if z < -1.96 or z > 1.96; (23 22)/(4/ 24) 1.22 z = = ; therefore, do
not reject H
0
. The data do not provide sufficient evidence to support
H
a
: =/ 22.
9.58 Reject H0 if z < -1.96 or z > 1.96; (20 22)/(4/ 24) 2.45 z = = ; therefore,
reject H0 and conclude that =/ 22.
9.59 n = 12, o = 0.37 ppm, x
_
= 6.31/12 = 0.526 ppm
Step 1: H
0
: = 0.5 ppm, H
a
: > 0.5 ppm
Step 2: o = 0.05
Step 3: 24 . 0 ) 12 / 37 . 0 /( ) 5 . 0 526 . 0 ( = = z
Step 4: Critical value = 1.645
Step 5: Since 0.24 < 1.645, do not reject H
0
.
Step 6: At the 5% significance level, the data do not provide sufficient
evidence to conclude that the mean cadmium level of Boletus
pinicola mushrooms is greater than the safety limit of 0.5 ppm.
9.60 n = 28, o = $8.45, x
_
= $1788.62/28 = $63.88
Step 1: H
0
: = $66.52, H
a
: =/ $66.52
Step 2: o = 0.10
Step 3: (63.88 66.52)/(8.45/ 28) 1.653 z = =
Step 4: Critical values = +1.645
Step 5: Since 1.653 < 1.645, reject H
0
.
Step 6: At the 10% significance level, the data do provide sufficient
evidence to conclude that the mean retail price of agricultural
books is different from the 2000 mean price.
9.61 n = 45, x
_
= 14.68, o = 4.2
Step 1: H0: = 18 mg, Ha: < 18 mg
Step 2: o = 0.01
Step 3: 30 . 5 ) 45 / 2 . 4 /( ) 18 68 . 14 ( = = z
Step 4: Critical value = -2.33
Step 5: Since -5.30 < -2.33, reject H0.
Step 6: At the 1% significance level, the data provide sufficient evidence
to conclude that adult females under the age of 51 are, on the
average, getting less than the RDA of 18 mg of iron. Considering
that iron deficiency causes anemia and that iron is required for
transporting oxygen in the blood, this result could have practical
significance as well.
9.62 n = 21, x
_
= 52.5 years, o = 6.8 years
Step 1: H
0
: = 55 years, H
a
: < 55 years
Step 2: o = 0.01
Step 3: (52.5 55)/(6.8/ 21) 1.68 z = =
Step 5: Since 1.68 > -2.33, do not reject H0.
evidence to conclude that the mean age of diagnosis of all people
with early-onset dementia is less than 55 years old.
9.63 n = 100, x
_
= 17.8 months, o = 6.0 months
Step 1: H
0
: = 16.7 months, H
a
: =/ 16.7 months
Step 2: o = 0.05
Step 3: 83 . 1 ) 100 / 0 . 6 /( ) 7 . 16 8 . 17 ( = = z
Step 4: Critical values = 1.96
Step 5: Since 1.96 < 1.83 < 1.96, do not reject H0.
evidence to conclude that the mean length of imprisonment of
motor-vehicle theft offenders in Sydney differs from the national
mean in Australia.
9.64 n = 29, x
_
= 78.3, o = 11.2
Step 1: H
0
: = 72 bpm, H
a
: > 72 bpm
Step 2: o = 0.05
Step 3: (78.3 72)/(11.2/ 29) 3.03 z = =
Step 5: Since 3.03 > 1.645, reject H0.
Step 6: At the 5% significance level, the data provide sufficient evidence
to conclude that the mean post-work heart rate for casting workings
exceeds the normal resting heart rate of 72 beats per minute.
GAIN
F
r
e
q
u
e
n
c
y
1. 0 0. 5 0. 0 -0.5 -1.0
5
4
3
2
1
0
_
X
Ho
Histogram of GAIN
(wi th Ho and 95% Z-c onfidence interval for the Mean, and S tDev = 0. 42)
GAIN
1. 0 0. 5 0. 0 -0. 5 -1. 0
_
X
Ho
Boxplotof GAIN
(with Ho and 95% Z-confidenc e interval for the Mean, and StDev = 0.42)
GAIN
P
e
r
c
e
n
t
1. 5 1. 0 0.5 0. 0 -0.5 -1.0
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
0.137
0.295
S tDev 0.5000
N 20
A D 0.549
P -V alue
Probability Plot of GAIN
Normal
9.65 (a) Using Minitab, with the data in a column named GAIN, we choose Stat

Basic Statistics

1-Sample z..., click in the Samples in columns text
box and specify GAIN, click in the Standard deviation text box and type
0.42, and click in the Test mean text box and type 0.2. Click the
Options button, enter 95 in the Confidence level text box, click the
arrow button at the right of the Alternative drop-down list box and
select greater than and click OK. Click on the Graphs button and check
the boxes for Histogram of Data and Boxplot of data. Then click OK
twice. The result of the test is
Test of mu = 0.2 vs > 0.2
The assumed standard deviation = 0.42
95%
Lower
Variable N Mean StDev SE Mean Bound Z P
GAIN 20 0.295000 0.499974 0.093915 0.140524 1.01 0.156
(b) The histogram and boxplot were produced by the procedure in part (a).
Now choose Stat

Basic Statistics

Normality test and enter GAIN in
the Variable text box. Click OK. Then choose Graph

Stem-and-Leaf ,
enter GAIN in the Graph variables text box and click OK. The results
are
Stem-and-leaf of GAIN N = 20
Leaf Unit = 0.10
LO -11
2 -0 5
3 -0 2
4 -0 1
6 0 01
10 0 2233
10 0 45
8 0 6667
4 0 8889
(c) Repeating the procedure of part (a), we obtain
CHARGE
F
r
e
q
u
e
n
c
y
120 100 80 60
5
4
3
2
1
0 _
X
Ho
Histogram of CHARGE
(with Ho and 95% Z-confidence interval for the Mean, and StDev = 22. 4)
CHARGE
130 120 110 100 90 80 70 60 50 40
_
X
Ho
Boxplotof CHARGE
(with Ho and 95% Z-confidenc e interval for the Mean, and StDev= 22.4)
95%
Lower
GAIN 19 0.368421 0.387374 0.096355 0.209932 1.75 0.040
(d) The original sample size is only 20. The plots in part (b) indicate
that the value 1.1 is a potential outlier. The z-test should not be
used with the original data. This is further confirmed by the fact
that using all of the data leads to z = 1.01, whereas, deleting the
outlier leads to z = 1.75. This is enough of a change to alter our
conclusion from not rejecting the null hypothesis to rejecting it. If
there is no good reason for deleting the outlier, then the z-test is
inappropriate for these data.

Basic Statistics

box and specify CHARGE, click in the Standard deviation text box and
type 22.4, and click in the Test mean text box and type 75. Click the
Test of mu = 75 vs < 75
95%
Upper
CHARGE 15 69.3560 24.3201 5.7837 78.8693 -0.98 0.165
Now choose Stat

Basic Statistics

Normality test and enter CHARGE
in the Variable text box. Click OK. Then choose Graph

Stem-and-
Leaf, enter CHARGE in the Graph variables text box and click OK. The
results are
CHARGE
P
e
r
c
e
n
t
140 120 100 80 60 40 20 0
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
0.009
69.36
StDev 24.32
N 15
AD 0.993
P- Valu e
Probability Plot of CHARGE
Normal
TEMP
F
r
e
q
u
e
n
c
y
99.5 99. 0 98.5 98. 0 97.5 97. 0
20
15
10
5
0
_
X
Ho
Histogram of TEMP
(wi th Ho and 99% Z-c onfidence interval for the Mean, and S tDev = 0. 63)
TEMP
99. 5 99. 0 98. 5 98. 0 97. 5 97. 0
_
X
Ho
Boxplotof TEMP
Stem-and-leaf of CHARGE N = 15
Leaf Unit = 1.0
2 4 77
(6) 5 013678
7 6 19
5 7 4
4 8 1
3 9 5
2 10 6
HI 130
(c) Repeating the procedure of part (a), we obtain
95%
Upper
CHARGE 14 65.0121 18.2251 5.9867 74.8593 -1.67 0.048
(d) The sample size is small, there is an outlier, and the data are skewed
right with and without the outlier 130.17. Therefore, the use of the
z-test is inappropriate for these data.
9.67 (a) Using Minitab, with the data in a column named TEMP, we choose Stat

Basic Statistics

box and specify TEMP, click in the Standard deviation text box and
enter 0.63, and click in the Test mean text box and enter 98.6. Click
the Options button, enter 95 in the Confidence level text box, click
the arrow button at the right of the Alternative drop-down list box and
select not equal and click OK. Click on the Graphs button and check
twice. Now choose Stat

Basic Statistics

Normality test and enter
CHARGE in the Variable text box. Click OK. Then choose Graph

Stem-
and-Leaf, enter CHARGE in the Graph variables text box and click OK.
The graphs are
TEMP
P
e
r
c
e
n
t
100 99 98 97 96
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
0.124
98.12
StDev 0.6468
N 93
A D 0.585
P- Value
Probability Plotof TEMP
Normal
Stem-and-leaf of TEMP N = 93
Leaf Unit = 0.10
1 96 7
3 96 89
8 97 00001
13 97 22233
19 97 444444
26 97 6666777
31 97 88889
45 98 00000000000111
(10) 98 2222222233
38 98 4444445555
28 98 66666666677
17 98 8888888
10 99 00001
5 99 2233
1 99 4
(b) Yes. The sample size 93 is large and the distribution of the data is
quite symmetric.
(c) Yes. The procedure in part (a) also produced the results of the test
which are
Test of mu = 98.6 vs not = 98.6
Variable N Mean StDev SE Mean 99% CI Z P
TEMP 93 98.1237 0.6468 0.0653 (97.9554, 98.2919) -7.29 0.000
The critical values are 2.575 and 2.575. Since z = -7.29, we reject
the null hypothesis and conclude that the mean body temperature of
healthy humans is different from the generally accepted value of
98.6F.
9.68 (a) Using Minitab, with the data in a column named SALARY, we choose Stat

Basic Statistics

box and specify SALARY, click in the Standard deviation text box and
select less than and click OK. Click on the Graphs button and check

Basic Statistics

SALARY in the Variable text box. Click OK. Then choose Graph

Stem-
and-Leaf, enter SALARY in the Graph variables text box and click OK.
The graphs are
SALARY
P
e
r
c
e
n
t
80 70 60 50 40 30 20 10
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
0.394
44.50
StDev 9.181
N 90
AD 0.381
P- Valu e
Probability Plot of SALARY
Normal
SALARY
F
r
e
q
u
e
n
c
y
64 56 48 40 32 24
16
12
8
4
0
_
X
Ho
Histogram of SALARY
SALARY
70 60 50 40 30 20
_
X
Ho
Boxplot of SALARY
(with Ho and 95% Z-confidence i nterval for the Mean, and StDev = 9. 2)
Stem-and-leaf of SALARY N = 90
Leaf Unit = 1.0
1 2 3
6 2 78889
18 3 012223444444
28 3 5666667899
44 4 0011111223344444
(20) 4 55566666777788889999
26 5 000001122223444
11 5 66688889
3 6 01
1 6 6
quite symmetric.
which are
Test of mu = 45.9 vs < 45.9
95%
Upper
SALARY 90 44.5033 9.1806 0.9698 46.0985 -1.44 0.075
The critical value is 1.645. Since z = -1.44, we do not reject the
null hypothesis. The data do not provide sufficient evidence that the
mean annual salary of classroom teachers in Hawaii is less than the
national mean.
9.69 (a) Using Minitab, with the data in a column named BILL, we choose Stat

Basic Statistics

box and specify BILL, click in the Standard deviation text box and
enter 25, and click in the Test mean text box and enter 47.37. Click

Basic Statistics

BILL
P
e
r
c
e
n
t
140 120 100 80 60 40 20 0 -20 -40
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
<0.005
50.64
StDev 23.75
N 75
AD 1.978
P- Value
Probability Plot of BILL
Normal
BILL
F
r
e
q
u
e
n
c
y
120 100 80 60 40 20
20
15
10
5
0 _
X
Ho
Histogramof BILL
(with Ho and 95%Z-confidence interval for the Mean, and StDev= 25)
BILL
120 100 80 60 40 20 0
_
X
Ho
Boxplot of BILL
(with Ho and 95% Z-c onfidence interval for the Mean, and StDev = 25)
BILL in the Variable text box. Click OK. Then choose Graph

Stem-
and-Leaf, enter BILL in the Graph variables text box and click OK. The
graphs are
Stem-and-leaf of BILL N = 75
Leaf Unit = 1.0
1 1 3
4 1 669
5 2 0
14 2 567777889
20 3 012334
31 3 55677889999
(7) 4 1122333
37 4 55678899
29 5 0011334
22 5 9
21 6 044
18 6 688
15 7 13
13 7
13 8 0112
9 8 7799
5 9 2
4 9 78
HI 114, 119
(b) The results of the test carried out by the procedure in part (a) are
Test of mu = 47.37 vs > 47.37
The assumed standard deviation = 25
95%
Lower
BILL 75 50.6405 23.7497 2.8868 45.8922 1.13 0.129
The critical value for the test is z = 1.645. Since z = 1.13, we do
not reject the null hypothesis. The data do not provide sufficient
evidence that the mean local monthly cell phone bill has increased from
the 2001 mean of $47.37.
After deleting the two outliers 119.61 and 114.98, the graphs and test
results are
BILL
P
e
r
c
e
n
t
120 90 60 30 0
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
<0.005
48.81
StDev 21.28
N 73
AD 1.688
P-V alu e
Normal
BILL
100 90 80 70 60 50 40 30 20 10
_
X
Ho
Boxplotof BILL
(with Ho and 95% Z-confidence interval for the Mean, and StDev= 25)
BILL
F
r
e
q
u
e
n
c
y
100 80 60 40 20
20
15
10
5
0
_
X
Ho
Histogramof BILL
(with Ho and 95% Z-confidenc e interval for the Mean, and StDev = 25)
Leaf Unit = 1.0
1 1 3
4 1 669
5 2 0
14 2 567777889
20 3 012334
31 3 55677889999
(7) 4 1122333
35 4 55678899
27 5 0011334
20 5 9
19 6 044
16 6 688
13 7 13
11 7
11 8 0112
7 8 7799
3 9 2
2 9 78
Test of mu = 47.37 vs > 47.37
95%
Lower
BILL 73 48.8144 21.2785 2.9260 44.0015 0.49 0.311
Although the value of z has been changed from 1.13 to 0.49 by deleting
the two outliers, the conclusion remains the same. We do not reject
the null hypothesis. Intuitively, we should expect this result, since
deleting two large outliers can only reduce the sample mean, making z
smaller.
9.70 (a) n = 28, o = $8.45, x
_
= $1788.62/28 = $63.88
The 90% confidence interval is 63.88 1.645(8.45)/ 28 (61.25,66.51) = .
The hypothesized mean ($66.52) lies outside of the confidence interval,
so we should reject the null hypothesis. Since the test statistic is
(63.88 66.52)/(8.45/ 28) 1.653 z = = , which is less than the lower
critical value of 1.645, the hypothesis test also leads to the
conclusion that we should reject the null hypothesis.
(b) n = 100, x
_
= 17.8 months, o = 6.0 months
The hypothesized mean (16.7) lies inside of the confidence interval, so
we should not reject the null hypothesis. Since the test statistic is
83 . 1 ) 100 / 0 . 6 /( ) 7 . 16 8 . 17 ( = = z , which is less than the upper critical
value of 1.96, the hypothesis test also leads to the conclusion that we
should not reject the null hypothesis.
9.71 (a) n = 45, x
_
= 14.68, o = 4.2
The 99% upper level confidence bound is 14.68 2.33(4.2)/ 45 16.14 + = .
The hypothesized mean (18 mg) lies above the upper confidence bound, so
we should reject the null hypothesis. Since the test statistic is
30 . 5 ) 45 / 2 . 4 /( ) 18 68 . 14 ( = = z , which is less than the critical value of
2.33, the hypothesis test also leads to the conclusion that we should
reject the null hypothesis.
(b) n = 21, x
_
= 52.5 years, o = 6.8 years
The 95% upper level confidence bound is 52.5 2.33(6.8)/ 21 55.96 + = .
The hypothesized mean (55) lies below the upper confidence bound, so we
should not reject the null hypothesis. Since the test statistic is
(52.5 55)/(6.8/ 21) 1.68 z = = , which is greater than the critical value
of 2.33, the hypothesis test also leads to the conclusion that we
9.72 (a) n = 12, o = 0.37 ppm, x
_
= 6.31/12 = 0.526 ppm
The 95% lower level confidence bound is 0.526 1.645(0.37)/ 12 0.350 = .
The hypothesized mean (0.5) lies above the lower confidence bound, so
24 . 0 ) 12 / 37 . 0 /( ) 5 . 0 526 . 0 ( = = z , which is less than the critical value of
not reject the null hypothesis.
(b) n = 30, x
_
= 78.3, o = 11.2
The 95% lower level confidence bound is 78.3 1.645(11.2)/ 30 74.94 = .
The hypothesized mean (72) lies below the lower confidence bound, so we
should reject the null hypothesis. Since the test statistic is
08 . 3 ) 30 / 2 . 11 /( ) 72 3 . 78 ( = = z , which is greater than the critical value
should reject the null hypothesis.
Exercises 9.4
9.73 Hypothesis tests have built-in margins of error. Errors will occur due to
the uncontrollable randomness in the data observed.
Section 9.4, Type II Error Probabilities; Power 397
9.74 (a) A Type I error occurs if the data leads to rejecting the null
hypothesis when it is, in fact, true.
(b) A Type II error occurs if the data leads to not rejecting the null
hypothesis when it is, in fact, false.
(c) The significance level is the probability associated with the test
procedure of rejecting the null hypothesis when it is actually true,
i.e., it is the probability of making a Type I error.
9.75 (a) = significance level = P(Type I error) = P(rejecting a true null
hypothesis)
(b) = P(Type II error) = P(not rejecting a false null hypothesis)
(c) 1 = Power of the test = P(rejecting a false null hypothesis)
9.76 The power of a hypothesis test is the probability of making the correct
decision of rejecting a false null hypothesis; that is, the probability of
not committing a Type II error. Power = 1 | .
9.77 Since is unknown, the power curve enables one to evaluate the
effectiveness of a hypothesis test for a variety of values of .
9.78 The power of a hypothesis test increases if the sample size is increased
without changing the significance level. This makes sense because larger
sample sizes should make it possible to detect smaller differences from the
null hypothesis (or make it more likely to detect a difference of a given
size).
9.79 If the significance level is decreased without changing the sample size, the
rejection region is made smaller (in probability terms). This makes the
nonrejection region larger, i.e., | gets larger. This, in turn, makes the
power 1 - | smaller.
9.80 Procedure B. Assuming that the assumptions underlying both procedures are
satisfied, we would choose the procedure with the greater power since it
gives us a better chance of rejecting a false null hypothesis.
9.81 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a right-tailed test, we would reject H
0
if z > 1.645; or
equivalently if 6757 . 0 12 / ) 37 . 0 ( 645 . 1 5 . 0 = + > x
So we reject H
0
if x
_
> 0.676; otherwise do not reject H
0
.
(b) P(Type I error) = o = 0.05
(c)
True mean z-score P(Type II error) Power
computation | 1 |
0.55 18 . 1
12 / 37 . 0
55 . 0 676 . 0
=
= z 0.8810 0.119 0
0.60 71 . 0
12 / 37 . 0
60 . 0 676 . 0
=
= z 0.7611 0.2389
0.65 24 . 0
12 / 37 . 0
65 . 0 676 . 0
=
= z 0.5948 0.4052
0.70 22 . 0
12 / 37 . 0
70 . 0 676 . 0
=
= z 0.4129 0.5871
0.75 69 . 0
12 / 37 . 0
75 . 0 676 . 0
=
= z 0.2451 0.7549
0.80 16 . 1
12 / 37 . 0
80 . 0 676 . 0
=
= z 0.1230 0.8770
0.85 63 . 1
12 / 37 . 0
85 . 0 676 . 0
=
= z 0.0516 0.9484
(d)
9.82 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a two-tailed test, we would reject H0 if |z| > 1.645; or
equivalently if 66.52 1.645(8.45)/ 28 69.15 x > + = or
66.52 1.645(8.45)/ 28 63.89 x s = .
So reject H0 if x
_
s 63.89 or x
_
> 69.15; otherwise do not reject H0.
PowerCurve
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
0.00 0.50 1.00
TrueMean
(c)
computation | 1 |
62
63.89 62
1.18
8.45/ 28
69.15 62
4.48
8.45/ 28
z
z
= =
= =
1.0000 0.8810 = 0.1190 0.8810
63
63.89 63
0.56
8.45/ 28
69.15 63
3.85
8.45/ 28
z
z
= =
= =
0.9999 0.7123 = 0.2876 0.7124
64
63.89 64
0.07
8.45/ 28
69.15 64
3.22
8.45/ 28
z
z
= =
= =
0.9994 0.4721 = 0.5273 0.4727
65
63.89 65
0.70
8.45/ 28
69.15 65
2.60
8.45/ 28
z
z
= =
= =
0.9953 0.2420 = 0.7533 0.2467
66
63.89 66
1.32
8.45/ 28
69.15 66
1.97
8.45/ 28
z
z
= =
= =
0.9756 0.0934 = 0.8822 0.1178
67
63.89 67
1.95
8.45/ 28
69.15 67
1.35
8.45/ 28
z
z
= =
= =
0.9115 0.0256 = 0.8859 0.1141
68
63.89 68
2.57
8.45/ 28
69.15 68
0.72
8.45/ 28
z
z
= =
= =
0.7642 0.0051 = 0.7591 0.2409
69
63.89 69
3.20
8.45/ 28
69.15 69
0.09
8.45/ 28
z
z
= =
= =
0.5374 0.0007 = 0.5367 0.4633
computation | 1 |
70
63.89 70
3.83
8.45/ 28
69.15 70
0.53
8.45/ 28
z
z
= =
= =
0.2981 0.0001 = 0.2980 0.7020
71
63.89 71
4.45
8.45/ 28
69.15 71
1.16
8.45/ 28
z
z
= =
= =
0.1230 0.0000 = 0.1230 0.8770
(d)
9.83 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a left-tailed test, we would reject H0 if z s -2.33; or
equivalently if 54 . 16 45 / ) 2 . 4 ( 33 . 2 18 = s x
So reject H0 if x
_
s 16.54; otherwise do not reject H0.
(c) Answers may differ slightly from those in the text due to intermediate
rounding.
computation | 1 - |
15.50 66 . 1
45 / 2 . 4
50 . 15 54 . 16
=
= z 1.000 0.9515 = 0.0485 0.9515

15.75 26 . 1
45 / 2 . 4
75 . 15 54 . 16
=
= z 1.000 0.8962 = 0.1038 0.8962

Power Curve
0.00
0.20
0.40
0.60
0.80
1.00
60 62 64 66 68 70 72
True Mean
P
o
w
e
r
computation | 1 - |
16.00 86 . 0
45 / 2 . 4
00 . 16 54 . 16
=
= z 1.000 0.8051 = 0.1949 0.8051

16.25 46 . 0
45 / 2 . 4
25 . 16 54 . 16
=
= z 1.000 0.6772 = 0.3228 0.6772

16.50 06 . 0
45 / 2 . 4
50 . 16 54 . 16
=
= z 1.000 0.5239 = 0.4761 0.5239

16.75 34 . 0
45 / 2 . 4
75 . 16 54 . 16
=
= z 1.000 0.3669 = 0.6331 0.3669

17.00 73 . 0
45 / 2 . 4
00 . 17 54 . 16
=
= z 1.000 0.2327 = 0.7673 0.2327

17.25 13 . 1
45 / 2 . 4
25 . 17 54 . 16
=
= z 1.000 0.1292 = 0.8708 0.1292

17.50 53 . 1
45 / 2 . 4
50 . 17 54 . 16
=
= z 1.000 0.0630 = 0.9370 0.0630

17.75 93 . 1
45 / 2 . 4
75 . 17 54 . 16
=
= z 1.000 0.0268 = 0.9732 0.0268

(d)
9.84 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a left-tailed test, we would reject H
0
if z s -2.33; or
equivalently if 55 2.33(6.8)/ 21 51.54 x s =
So reject H
0
if x
_
s 51.54; otherwise do not reject H
0
.
PowerCurve
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
15 16 17 18
TrueMean
rounding.
computation | 1 - |
47
51.54 47
3.06
6.8/ 21
z
= = 1.000 0.9989 = 0.0011 0.9989

48
51.54 48
2.39
6.8/ 21
z
= = 1.000 0.9916 = 0.0084 0.9916

49
51.54 49
1.71
6.8/ 21
z
= = 1.000 0.9964 = 0.0036 0.9964

50
51.54 50
1.04
6.8/ 21
z
= = 1.000 0.8508 = 0.1492 0.8508

51
51.54 51
0.36
6.8/ 21
z
= = 1.000 0.6406 = 0.3594 0.6406

52
51.54 52
0.31
6.8/ 21
z
= = 1.000 0.3783 = 0.6217 0.3783

53
51.54 53
0.98
6.8/ 21
z
= = 1.000 0.1635 = 0.8365 0.1635

54
51.54 54
1.66
6.8/ 21
z
= = 1.000 0.0485 = 0.9515 0.0485

(d)
9.85 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a two-tailed test, we would reject H
0
if |z| > 1.96; or
equivalently if 876 . 17 100 / ) 0 . 6 ( 96 . 1 7 . 16 = + > x or
524 . 15 100 / ) 0 . 6 ( 96 . 1 7 . 16 = s x .
So reject H0 if x
_
s 15.524 or x
_
Power Curve
0.00
0.20
0.40
0.60
0.80
1.00
46 47 48 49 50 51 52 53 54 55
True Mean
P
o
w
e
r
(c)
computation | 1 |
14.0
46 . 6
100 / 0 . 6
0 . 14 876 . 17
54 . 2
100 / 0 . 6
0 . 14 524 . 15
=
=
=
=
z
z
1.0000 0.9945 = 0.0055 0.9945
14.5
63 . 5
100 / 0 . 6
5 . 14 876 . 17
71 . 1
100 / 0 . 6
5 . 14 524 . 15
=
=
=
=
z
z
1.0000 0.9564 = 0.0436 0.9564
15.0
79 . 4
100 / 0 . 6
0 . 15 876 . 17
87 . 0
100 / 0 . 6
0 . 15 524 . 15
=
=
=
=
z
z
1.0000 0.8078 = 0.1922 0.8078
15.5
96 . 3
100 / 0 . 6
5 . 15 876 . 17
04 . 0
100 / 0 . 6
5 . 15 524 . 15
=
=
=
=
z
z
1.0000 0.5160 = 0.4540 0.5160
16.0
13 . 3
100 / 0 . 6
0 . 16 876 . 17
79 . 0
100 / 0 . 6
0 . 16 524 . 15
=
=
=
=
z
z
0.9991 0.2148 = 0.7843 0.2157
16.5
29 . 2
100 / 0 . 6
5 . 16 876 . 17
63 . 1
100 / 0 . 6
5 . 16 524 . 15
=
=
=
=
z
z
0.9890 0.0516 = 0.9374 0.0626
17.0
46 . 1
100 / 0 . 6
0 . 17 876 . 17
46 . 2
100 / 0 . 6
0 . 17 524 . 15
=
=
=
=
z
z
0.9279 0.0069 = 0.9210 0.0790
17.5
63 . 0
100 / 0 . 6
5 . 17 876 . 17
29 . 3
100 / 0 . 6
5 . 17 524 . 15
=
=
=
=
z
z
0.7357 0.0005 = 0.7352 0.2648
computation | 1 |
18.0
21 . 0
100 / 0 . 6
0 . 18 876 . 17
13 . 4
100 / 0 . 6
0 . 18 524 . 15
=
=
=
=
z
z
0.4168 0.0000 = 0.4168 0.5832
18.5
04 . 1
100 / 0 . 6
5 . 18 876 . 17
96 . 4
100 / 0 . 6
5 . 18 524 . 15
=
=
=
=
z
z
0.1492 0.0000 = 0.1492 0.8508
19.0
87 . 1
100 / 0 . 6
0 . 19 876 . 17
79 . 5
100 / 0 . 6
0 . 19 524 . 15
=
=
=
=
z
z
0.0307 0.0000 = 0.0307 0.9693
(d)
9.86 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
0
if z > 1.645; or
equivalently if 42 . 75 29 / ) 2 . 11 ( 645 . 1 72 = + > x
So reject H
0
if x
_
0
.
rounding.
Power Curve
0.00
0.20
0.40
0.60
0.80
1.00
13 14 15 16 17 18 19 20
True Mean
P
o
w
e
r
Power Curve
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
72 74 76 78 80 82
True Mean
P
o
w
e
r
computation | 1 |
73 16 . 1
29 / 2 . 11
) 73 42 . 75 (
=
= z 0.8777 0.1223
74 68 . 0
29 / 2 . 11
) 74 42 . 75 (
=
= z 0.7526 0.2474
75 20 . 0
29 / 2 . 11
) 75 4 . 75 (
=
= z 0.5800 0.4200
76 28 . 0
29 / 2 . 11
) 76 4 . 75 (
=
= z 0.3902 0.6098
77 76 . 0
29 / 2 . 11
) 77 4 . 75 (
=
= z 0.2237 0.7763
78 24 . 1
29 / 2 . 11
) 78 4 . 75 (
=
= z 0.1074 0.8926
79 72 . 1
29 / 2 . 11
) 79 4 . 75 (
=
= z 0.0426 0.9574
80 20 . 2
29 / 2 . 11
) 80 4 . 75 (
=
= z 0.0138 0.9862
(d)
9.87 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a right-tailed test, we would reject H0 if z > 1.645; or
So we reject H0 if x
_
rounding.
computation | 1 |
0.55 04 . 1
20 / 37 . 0
55 . 0 636 . 0
=
= z 0.8508 0.1492
0.60 44 . 0
20 / 37 . 0
60 . 0 636 . 0
=
= z 0.6700 0.3300
0.65 17 . 0
20 / 37 . 0
65 . 0 636 . 0
=
= z 0.4325 0.5675
0.70 77 . 0
20 / 37 . 0
70 . 0 636 . 0
=
= z 0.2206 0.7794
0.75 38 . 1
20 / 37 . 0
75 . 0 636 . 0
=
= z 0.0838 0.9162
0.80 98 . 1
20 / 37 . 0
80 . 0 636 . 0
=
= z 0.0239 0.9761
0.85 59 . 2
20 / 37 . 0
85 . 0 636 . 0
=
= z 0.0048 0.9952
(d)
The power curve with n = 20 rises more quickly as the true mean
increases, resulting in a higher power at any given value of than
for n = 12. This illustrates the principle that a larger sample size
has a higher probability of rejecting the null hypothesis when the
null hypothesis is false and the significance level remains the same.
9.88 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
PowerCurve
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
0.00 0.50 1.00
TrueMean
66.52 1.645(8.45)/ 50 64.55 x s = .
So reject H0 if x
_
s 64.55 or x
_
(c) The details of the z-score computation are the same as in Exercise
9.82, with 64.55 replacing 63.89, 68.49 replacing 69.15, and 50
replacing 28. The results of the computations are
P(Type II) Power
z-left z-right P(z<z-left) P(z<z-rt) 1-
62 1.60 4.06 0.9452 1.0000 0.0548 0.9452
63 0.97 3.44 0.8340 0.9997 0.1657 0.8343
64 0.34 2.81 0.6331 0.9975 0.3645 0.6355
65 -0.28 2.19 0.3897 0.9857 0.5960 0.4040
66 -0.91 1.56 0.1814 0.9406 0.7592 0.2408
67 -1.53 0.93 0.0630 0.8238 0.7608 0.2392
68 -2.16 0.31 0.0154 0.6217 0.6063 0.3937
69 -2.79 -0.32 0.0026 0.3745 0.3718 0.6282
70 -3.41 -0.95 0.0003 0.1711 0.1707 0.8293
71 -4.04 -1.57 0.0000 0.0582 0.0582 0.9418
(d)
The power curve with n = 50 rises more quickly as the true mean
increases or decreases from $66.52, resulting in a higher power at any
given value of than for n = 28. This illustrates the principle
that a larger sample size has a higher probability of rejecting the
null hypothesis when the null hypothesis is false and the significance
level remains the same.
9.89 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
16.7 1.96(6.0)/ 40 14.841 x s = .
So reject H0 if x
_
s 14.841 or x
_
Power Curve
0.00
0.20
0.40
0.60
0.80
1.00
60 62 64 66 68 70 72
True Mean
P
o
w
e
r
9.85, with 14.841 replacing 15.524, 18.559 replacing 17.876, and 40
replacing 100. The results of the computations are
Power
z-left z-right P(z<z-left)P(z<z-rt) 1-
14.0 0.89 4.81 0.8133 1.0000 0.1867 0.8133
14.5 0.36 4.28 0.6406 1.0000 0.3594 0.6406
15.0 -0.17 3.75 0.4325 0.9999 0.5674 0.4326
15.5 -0.69 3.22 0.2451 0.9994 0.7543 0.2457
16.0 -1.22 2.70 0.1112 0.9965 0.8853 0.1147
16.5 -1.75 2.17 0.0401 0.9850 0.9449 0.0551
17.0 -2.28 1.64 0.0113 0.9495 0.9382 0.0618
17.5 -2.80 1.12 0.0026 0.8686 0.8661 0.1339
18.0 -3.33 0.59 0.0004 0.7224 0.7220 0.2780
18.5 -3.86 0.06 0.0001 0.5239 0.5239 0.4761
19.0 -4.38 -0.46 0.0000 0.3228 0.3228 0.6772
The power curve with n = 40 rises less quickly as the true mean
increases or decreases from 16.7, resulting in a lower power at any
given value of than for n = 100. This illustrates the principle
that a larger sample size has a higher probability of rejecting the
null hypothesis when the null hypothesis is false and the significance
level remains the same.
9.90 Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a left-tailed test, we would reject H0 if z s -2.33; or
equivalently if 55 2.33(6.8)/ 15 50.91 x s =
So reject H0 if x
_
s 50.91; otherwise do not reject H0.
9.85, with 50.91 replacing 51.54 and 15 replacing 21. The results of
the computations are
Power Curve
0.00
0.20
0.40
0.60
0.80
1.00
13 14 15 16 17 18 19 20
True Mean
P
o
w
e
r
Power
1
0
Power
1
0
P(Type
II) Power
z P(Z<z) 1-
47 2.63 0.9957 0.0043 0.9957
48 1.96 0.9750 0.0250 0.9750
49 1.29 0.9015 0.0985 0.9015
50 0.61 0.7291 0.2709 0.7291
51 -0.06 0.4761 0.5239 0.4761
52 -0.73 0.2327 0.7673 0.2327
53 -1.41 0.0793 0.9207 0.0793
54 -2.08 0.0188 0.9812 0.0188
(d)
The power curve with n = 15 rises less quickly as the true mean or
decreases from 55, resulting in a lower power at any given value of
than for n = 21. This illustrates the principle that a larger sample
size has a higher probability of rejecting the null hypothesis when
the null hypothesis is false and the significance level remains the
same.
9.91 (a)
(b) The curve in part (a) portrays that, ideally, one desires the value
for the power for any given
a
to be as close to 1 as possible.
9.92 (a)
Power Curve
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
46 47 48 49 50 51 52 53 54 55
True Mean
P
o
w
e
r
Power
1
0
(b) The curve in part (a) portrays that, ideally, one desires the value
for the power for any given
a
9.93 (a)
(b) The curve in (a) portrays that, ideally, one desires the value for the
power for any given
a
9.94 (a) Note:
0
0
/
/
x
z x z n
n
o
o
= = +
Since this is a left-tailed test, reject H0 if z < -1.645; or
equivalently if 6 . 25 30 / ) 4 . 1 ( 645 . 1 26 = s x
So reject H0 if x
_
s 25.6; otherwise do not reject H0. If is really
25.4, then P(Type II error) = P(x
_
> 25.6) =
25.6 25.4
( ) ( 0.78) 1.0000 0.7823 0.2177.
1.4/ 30
P z P z
> = > = =
(b)-(c) Answers will vary. You can use the procedure outlined in the
solution to Exercise 9.43 to generate the samples using the Excel
spreadsheet.
(d) Since the probability of a Type II error when = 25.4 (i.e., of
failing to reject the null hypothesis when = 25.4) is 0.2177 from
part (a), we would expect non-rejection of the null hypothesis about 21
or 22 times in the 100 samples.
(e)-(f) Answers will vary. Our simulation contained 20 samples in which
the null hypothesis was not rejected when =25.4. This is very close to
what we expected.
Exercises 9.5
9.95 (1) It allows the reader to assess significance at any desired level, and
(2) it permits the reader to evaluate the strength of the evidence against
9.96 The P-value of a test is, assuming the null hypothesis to be true, the
probability of observing a value of the test statistic as extreme or more
extreme than the one that was actually observed. When the P-value is small,
it provides evidence against the null hypothesis.
9.97 In the critical value approach, we determine critical values based on the
significance level. The critical values determine where the rejection and
nonrejection regions lie for the test statistic. If the value of the test
statistic falls in the rejection region, the null hypothesis is rejected. In
the P-value approach, the test statistic is computed and then the
probability of observing a value as extreme or more extreme than the value
obtained is determined. If the P-value is smaller than the significance
Section 9.5, P-Values 411
level, the null hypothesis is rejected. Reporting a P-value allows a reader
to draw his/her own conclusion based on the strength of the evidence.
9.98 The P-value for a one-sample z-test is obtained as:
(a) P(z < observed z value) for a left-tailed test
(b) P(z > observed z value) for a right-tailed test
(c) 2P(z > absolute value of the observed z value) for a two-tailed test.
9.99 True
9.100 (a) Do not reject the null hypothesis.
(b) Reject the null hypothesis.
(c) Reject the null hypothesis.
9.101 (a) Do not reject the null hypothesis.
(b) Reject the null hypothesis.
(c) Do not reject the null hypothesis.
9.102 A P-value of 0.02 provides stronger evidence against the null hypothesis
than does a value of 0.03. It says that if the null hypothesis is true, the
data are less likely than they are when the P-value is 0.03.
9.103 (a) Strength of the evidence against the null hypothesis is moderate.
(b) There is weak or no evidence against the null hypothesis.
(c) Strength of the evidence against the null hypothesis is strong.
(d) Strength of the evidence against the null hypothesis is very strong.
9.104 (a) Strength of the evidence against the null hypothesis is weak or none.
(b) Strength of the evidence against the null hypothesis is moderate.
(c) Strength of the evidence against the null hypothesis is very strong.
(d) Strength of the evidence against the null hypothesis is strong.
9.105 (a) ( )/( / ) (20 22)/(4/ 32) 2.83 z x n o = = = ; P-value = P(z < -2.83) = 0.0023
(b) The evidence against the null hypothesis is very strong.
9.106 (a) ( )/( / ) (21 22)/(4/ 32) 1.41 z x n o = = = ; P-value = P(z < -1.41) = 0.0793
(b) The evidence against the null hypothesis is moderate.
9.107 (a) ( )/( / ) (24 22)/(4/ 15) 1.94 z x n o = = = ; P-value = P(z > 1.94) = 0.0262
(b) The evidence against the null hypothesis is strong.
9.108 (a) ( )/( / ) (23 22)/(4/ 15) 0.97 z x n o = = = ; P-value = P(z > 0.97) = 0.1660
(b) The evidence against the null hypothesis is weak or none.
9.109 (a) ( )/( / ) (23 22)/(4/ 24) 1.22 z x n o = = = ; P-value = 2P(z > |1.22|) = 0.2224
(b) The evidence against the null hypothesis is weak or none.
9.110 (a) ( )/( / ) (20 22)/(4/ 24) 2.45 z x n o = = = ; P-value = 2P(z > |-2.45|) = 0.0142
(b) The evidence against the null hypothesis is strong.
9.111 (a) z = 2.03, P-value = 1.0000 - 0.9788 = 0.0212
(b) z = -0.31, P-value = 1.0000 - 0.3783 = 0.6217
9.112 (a) z = -1.84, P-value = 0.0329
(b) z = 1.25, P-value = 0.8944
9.113 (a) z = -0.74, P-value = 0.8770
(b) z = 1.16, P-value = 0.0329
9.114 (a) z = 3.08, Right-tail probability = 1.0000 - 0.9990 = 0.0010
P-value = 0.001 x 2 = 0.0020
(b) z = -2.42, Left-tail probability = 0.0078
P-value = 0.0078 x 2 = 0.0156
9.115 (a) z = -1.66, Left-tail probability = 0.0485
P-value = 0.0485 x 2 = 0.0970
(b) z = 0.52, Right-tail probability = 1.0000 - 0.6985 = 0.3015
P-value = 0.3015 x 2 = 0.6030
9.116 (a) z = 1.24, P-value = 1.0000 - 0.8925 = 0.1075
(b) z = -0.69, P-value = 1.0000 - 0.2451 = 0.7549
9.117 (See Exercise 9.59 for classical approach results.)
Step 1: H
0
: = 0.5 ppm, H
a
: > 0.5 ppm
Step 2: o = 0.05
Step 3: z = 0.24
Step 4: P-value = P(z 0.24) = 1.0000 - 0.5948 = 0.4052
Step 5: Since 0.4052 > 0.05, do not reject H
0
.
evidence to conclude that the mean cadmium level of Boletus
pinicola mushrooms is greater than the safety limit of 0.5 ppm.
Using Table 9.12, we classify the strength of evidence against the null
hypothesis as weak or none because P > 0.10.
Step 1: H
0
: = $66.52, H
a
: $66.52
Step 2: o = 0.10
Step 3: z = -1.65
Step 4: P-value = 2P(z 1.65) = 2(1.0000 - 0.9505) = 0.0990
0
.
evidence to conclude that the mean retail price of agriculture
books has changed from the 2000 mean.
hypothesis as moderate because 0.05 < P < 0.10.
Step 1: H0: = 18 mg, Ha: < 18 mg
Step 2: o = 0.01
Step 3: z = -5.30
Step 4: P-value = P(z < -5.30) = 0.0000 (to four decimal places)
Step 5: Since 0.0000 < 0.01, reject H0.
Step 6: At the 1% significance level, the data provide sufficient
evidence to conclude that adult females under the age of 51 are,
on the average, getting less than the RDA of 18 mg of iron.
hypothesis as very strong because P < 0.01.
Step 1: H0: = 55 years, Ha: < 55 years
Step 2: o = 0.01
Step 3: z = -1.69
Step 4: P-value = P(z < -1.69) = 0.0455
0
.
evidence to conclude that the mean age at diagnosis of all
people with early-onset dementia is less than 55 years old.
hypothesis as strong because 0.01 1.83) = 2(0.0336) =0.0672
Step 5: Since 0.0672 > 0.05, do not reject H0.
evidence to conclude that the mean length of imprisonment for
motor-vehicle theft offenders in Sydney, Australia differs from
the Australian national mean of 16.7 months.
hypothesis as moderate since 0.05 72
Step 2: o = 0.05
Step 3: z = 3.08
Step 4: P-value = P(z > 3.08) = 1.0000 - 0.9990 = 0.0010
0
.
evidence to conclude that the post-work heart rate for casting
workers is greater than the normal resting heart rate of 72 bpm.
hypothesis as very strong because P < 0.01.

Basic Statistics

box and specify GAIN, click in the Standard deviation text box and type
0.42, and click in the Test mean text box and type 0.2. Click the
GAIN
F
r
e
q
u
e
n
c
y
1. 0 0. 5 0. 0 -0.5 -1.0
5
4
3
2
1
0
_
X
Ho
Histogram of GAIN
(with Ho and 95% Z-c onfidence interval for the Mean, and StDev = 0. 42)
GAIN
1.0 0.5 0. 0 -0. 5 -1. 0
_
X
Ho
Boxplot of GAIN
(with Ho and 95% Z-confidenc e interval for the Mean, and StDev= 0. 42)
GAIN
P
e
r
c
e
n
t
1. 5 1. 0 0.5 0. 0 -0. 5 -1.0
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
0.137
0.295
StDev 0. 5000
N 20
AD 0.549
P-V alu e
Probability Plot of GAIN
Normal
95%
Lower
GAIN 20 0.295000 0.499974 0.093915 0.140524 1.01 0.156
The P-value is reported in the last line of the output as 0.156. Since
0.156 > 0.05, we do not reject the null hypothesis. The evidence
against the null hypothesis is weak or nonexistent.
Now choose Stat

Basic Statistics

Normality test and enter GAIN in
the Variable text box. Click OK. Then choose Graph

Stem-and-Leaf ,
enter GAIN in the Graph variables text box and click OK. The results
are
Stem-and-leaf of GAIN N = 20
Leaf Unit = 0.10
LO -11
2 -0 5
3 -0 2
4 -0 1
6 0 01
10 0 2233
10 0 45
8 0 6667
4 0 8889
(c) Repeating the procedure of part (a) after removing the outlier, we
obtain
95%
Lower
GAIN 19 0.368421 0.387374 0.096355 0.209932 1.75 0.040
Now the P-value is 0.04, which is less than the significance level,
leading to rejection of the null hypothesis. Now the data do provide
CHARGE
F
r
e
q
u
e
n
c
y
120 100 80 60
5
4
3
2
1
0 _
X
Ho
Histogram of CHARGE
(with Ho and 95% Z-confidence interval for the Mean, and StDev = 22. 4)
CHARGE
130 120 110 100 90 80 70 60 50 40
_
X
Ho
Boxplot of CHARGE
sufficient evidence to conclude that, on average, the net percentage
gain exceeds 0.2.
(d) The original sample size is only 20. The plots in part (b) indicate
that the value 1.1 is a potential outlier. The z-test should not be
used with the original data. This is further confirmed by the fact
that using all of the data leads to z = 1.01, whereas, deleting the
outlier leads to z = 1.75. This is enough of a change to alter our
conclusion from not rejecting the null hypothesis to rejecting it. If
there is no good reason for deleting the outlier, then the z-test is
inappropriate for these data.

Basic Statistics

box and specify CHARGE, click in the Standard deviation text box and
type 22.4, and click in the Test mean text box and type 75. Click the
95%
Upper
CHARGE 15 69.3560 24.3201 5.7837 78.8693 -0.98 0.165
The P-value is reported in the last line of the output as 0.165. Since
0.165 > 0.05, we do not reject the null hypothesis. The evidence
against the null hypothesis is weak or nonexistent.
Now choose Stat

Basic Statistics

Normality test and enter CHARGE
in the Variable text box. Click OK. Then choose Graph

Stem-and-
Leaf, enter CHARGE in the Graph variables text box and click OK. The
results are
CHARGE
P
e
r
c
e
n
t
140 120 100 80 60 40 20 0
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
0.009
69.36
StDev 24.32
N 15
AD 0.993
P- Value
Probability Plot of CHARGE
Normal
TEMP
F
r
e
q
u
e
n
c
y
99.5 99.0 98.5 98.0 97.5 97.0
20
15
10
5
0 _
X
Ho
Histogramof TEMP
(with Ho and 95% Z-confidence interval for the Mean, and StDev = 0.63)
TEMP
99.5 99.0 98.5 98.0 97.5 97.0
_
X
Ho
Boxplot of TEMP
(wi th Ho and 95%Z-confidence interval for the Mean, and StDev = 0.63)
Stem-and-leaf of CHARGE N = 15
Leaf Unit = 1.0
2 4 77
(6) 5 013678
7 6 19
5 7 4
4 8 1
3 9 5
2 10 6
HI 130
(c) Repeating the procedure of part (a) after deleting the outlier, we
obtain
95%
Upper
CHARGE 14 65.0121 18.2251 5.9867 74.8593 -1.67 0.048
The P-value is now 0.048, which is less than the significance level of
0.05, leading to rejection of the null hypothesis.
(d) The sample size is small, there is an outlier, and the data are skewed
right with and without the outlier 130.17. Therefore, the use of the
z-test is inappropriate for these data.
9.125 (a) Using Minitab, with the data in a column named TEMP, we choose Stat

Basic Statistics

box and specify TEMP, click in the Standard deviation text box and
select not equal and click OK. Click on the Graphs button and check

Basic Statistics

TEMP in the Variable text box. Click OK. Then choose Graph

Stem-
and-Leaf, enter TEMP in the Graph variables text box and click OK. The
graphs are
TEMP
P
e
r
c
e
n
t
100 99 98 97 96
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
0.124
98.12
StDev 0.6468
N 93
A D 0.585
P- Value
Probability Plotof TEMP
Normal
Stem-and-leaf of TEMP N = 93
Leaf Unit = 0.10
1 96 7
3 96 89
8 97 00001
13 97 22233
19 97 444444
26 97 6666777
31 97 88889
45 98 00000000000111
(10) 98 2222222233
38 98 4444445555
28 98 66666666677
17 98 8888888
10 99 00001
5 99 2233
1 99 4
quite symmetric.
which are
Variable N Mean StDev SE Mean 99% CI Z P
TEMP 93 98.1237 0.6468 0.0653 (97.9554, 98.2919) -7.29 0.000
The P-value is shown in the last line of the output as 0.000. This is
less than the significance level of 0.01, leading to rejection of the
null hypothesis. The evidence against the null hypothesis is
categorized as very strong since P < 0.01.
9.126 (a) Using Minitab, with the data in a column named SALARY, we choose Stat

Basic Statistics

box and specify SALARY, click in the Standard deviation text box and
select less than and click OK. Click on the Graphs button and check

Basic Statistics

SALARY in the Variable text box. Click OK. Then choose Graph

Stem-
and-Leaf, enter SALARY in the Graph variables text box and click OK.
The graphs are
SALARY
P
e
r
c
e
n
t
80 70 60 50 40 30 20 10
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
0.394
44.50
StDev 9.181
N 90
AD 0.381
P- Value
Probability Plot of SALARY
Normal
SALARY
F
r
e
q
u
e
n
c
y
64 56 48 40 32 24
16
12
8
4
0
_
X
Ho
Histogram of SALARY
SALARY
70 60 50 40 30 20
_
X
Ho
Boxplot of SALARY
(with Ho and 95% Z-confidence interval for the Mean, and StDev = 9.2)
Stem-and-leaf of SALARY N = 90
Leaf Unit = 1.0
1 2 3
6 2 78889
18 3 012223444444
28 3 5666667899
44 4 0011111223344444
(20) 4 55566666777788889999
26 5 000001122223444
11 5 66688889
3 6 01
1 6 6
quite symmetric.
which are
Test of mu = 45.9 vs < 45.9
95%
Upper
SALARY 90 44.5033 9.1806 0.9698 46.0985 -1.44 0.075
The P-value in the last line is 0.075 which is greater than the
significance level of 0.05. Therefore, we do not reject the null
hypothesis. The data do not provide sufficient evidence to conclude
that the mean teacher salary in Hawaii is less than the national
average.
9.127 (a) Using Minitab, with the data in a column named BILL, we choose Stat

Basic Statistics

box and specify BILL, click in the Standard deviation text box and
enter 25, and click in the Test mean text box and enter 47.37. Click

Basic Statistics

BILL in the Variable text box. Click OK. Then choose Graph

Stem-
and-Leaf, enter BILL in the Graph variables text box and click OK. The
graphs are
BILL
P
e
r
c
e
n
t
140 120 100 80 60 40 20 0 -20 -40
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
<0.005
50.64
StDev 23.75
N 75
AD 1.978
P-V alu e
Normal
BILL
120 100 80 60 40 20 0
_
X
Ho
Boxplot of BILL
(with Ho and 95% Z-confidence interval for the Mean, and StDev = 25)
BILL
F
r
e
q
u
e
n
c
y
120 100 80 60 40 20
20
15
10
5
0
_
X
Ho
Histogramof BILL
(with Ho and 95% Z-confidenc e interval for the Mean, and StDev= 25)
Leaf Unit = 1.0
1 1 3
4 1 669
5 2 0
14 2 567777889
20 3 012334
31 3 55677889999
(7) 4 1122333
37 4 55678899
29 5 0011334
22 5 9
21 6 044
18 6 688
15 7 13
13 7
13 8 0112
9 8 7799
5 9 2
4 9 78
HI 114, 119
(b) The results of the test carried out by the procedure in part (a) are
Test of mu = 47.37 vs > 47.37
95%
Lower
BILL 75 50.6405 23.7497 2.8868 45.8922 1.13 0.129
The P-value is shown in the last line as 0.129, which is greater than the
significance level of 0.05, leading to nonrejection of the null hypothesis.
The data do not provide sufficient evidence to conclude that the mean local
monthly cell phone bill has increased from the 2001 mean of $47.37.
The evidence against the null hypothesis is weak or nonexistent.
After deleting the two outliers 119.61 and 114.98, the graphs and test
results are
BILL
P
e
r
c
e
n
t
120 90 60 30 0
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
<0.005
48.81
StDev 21.28
N 73
AD 1.688
P- Value
Normal
BILL
100 90 80 70 60 50 40 30 20 10
_
X
Ho
Boxplot of BILL
(wi th Ho and 95% Z-c onfi dence interval for the Mean, and StDev = 25)
BILL
F
r
e
q
u
e
n
c
y
100 80 60 40 20
20
15
10
5
0
_
X
Ho
Histogram of BILL
Leaf Unit = 1.0
1 1 3
4 1 669
5 2 0
14 2 567777889
20 3 012334
31 3 55677889999
(7) 4 1122333
35 4 55678899
27 5 0011334
20 5 9
19 6 044
16 6 688
13 7 13
11 7
11 8 0112
7 8 7799
3 9 2
2 9 78
Test of mu = 47.37 vs > 47.37
95%
Lower
BILL 73 48.8144 21.2785 2.9260 44.0015 0.49 0.311
Although the P-value has been changed from 0.129 to 0.311 by deleting
the two outliers, the conclusion remains the same. We do not reject
the null hypothesis. Intuitively, we should expect this result, since
deleting two large outliers can only reduce the sample mean, making the
P-value larger.
9.128 (a) The P-value is expressed as P(z < z0) if the hypothesis test is left-
tailed.
(b) The P-value is expressed as P(z > z0) if the test is two-tailed.
9.129 (a) Left-tailed: P-value = P(z < z0) = (z0)
(b) Right-tailed: P-value = P(z > z0) = 1 - P(z < z0) = 1 - (z0)
(c) Two-tailed: P-value = P(z > z
0
)
= P(z > z
0
) + P(z < - z
0
)
Section 9.6, Hyp. Tests for One Pop. Mean When Is Unknown 421
By symmetry = P(z > z
0
) + P(z > z
0
)
= 2 P(z > z0) = 2 [1 - P(z < z0)]
= 2 [1 - (z0)]
9.130 Given that x can be transformed to z (and x0 to z0), we have:
1. P(x > x0) = P(z > z0), for a right-tailed test
2. P(x < x
0
) = P(z < z
0
), for a left-tailed test
3. 2 min {P(x < x
0
), P(x > x
0
)} = 2min {P(z < z
0
), P(z > z
0
)}
=
s <
> >
`
)
2 0
2 0
0 0
0 0
Pz z
Pz z
( )
( )
if z
if z
By symmetry =
> <
s >
`
)
2 0
2 0
0
0
Pz z
P z z
( )
( )
if z
if z
0
0
= 2P(z > z
0
)
By symmetry = P(z < -z
0
) + P(z > z
0
) = P(z > z
0
)
9.131 The P-value approach provides the actual significance of the hypothesis
test, that is, the smallest significance level at which the results of the
test are significant. It also allows the reader to judge the strength of
the evidence against the null hypothesis for himself/herself. The only
disadvantage of the P-value approach is that one must compute the P-value
(one extra step) after computing the z-value.
Exercises 9.6
9.132 In the z-test, it is assumed that o is known. In the t-test, o is
unknown.
9.133 (a) 0.01 0.025; Do not reject H0 for < 0.01; Undecided for
0.01 < < 0.025.
9.134 (a) P > 0.10
(b) Do not reject H0 for < 0.10; Undecided for > 0.10.
9.135 (a) P < 0.005
(b) Reject H
0
for > 0.005; Undecided for < 0.005.
9.136 (a) 0.05 0.01; Undecided for < 0.01.
9.139 (a) df = 31; t = -2.828
(b) P < 0.005; Reject H
0
; Evidence against H
0
is very strong.
9.140 (a) df = 31; t = -1.414
(b) 0.05 < P < 0.10; Do not reject H0; Evidence against H0 is moderate.
9.141 (a) df = 14; t = 1.936
(b) 0.05 0.20; Do not reject H0; Evidence against H0 is weak or none.
9.144 (a) df = 23; t = -2.449
(b) 0.02 1.729, we reject H
0
. The data provide sufficient evidence to
conclude that the amount of television watched per day last year by the
average person differed from that in 2002.
9.146 H
0: = 180 yards, Ha: > 180 yards, = 0.05;
( )/( / ) (182.7 180)/(2.7/ 6) 2.449 t x n o = = = ; 0.025 < P < 0.05
Since P-value < 0.05, we reject H
0
. The data provide sufficient evidence to
conclude that the club will hit the ball more than 180 years at a club head
velocity of 85 mph.
9.147 n = 10, df = 9, x
_
= 2.5, s = 0.149
Step 1: H0: = 2.3, Ha: > 2.3
Step 2: o = 0.01
Step 3: 251 . 4
10 / 149 . 0
3 . 2 5 . 2
=
= t
Step 5: Since 4.251 > 2.821, reject H0. Note: For the P-value approach,
P-value < 0.01. So, since the P-value < o , reject H0.
evidence to conclude that the mean available limestone in soil
treated with 100% MMBL effluent is greater than 2.30%. The
practical significance of this result probably depends on what crop
is to be grown in the soil.
9.148 n = 25, df = 24, x
_
= $1935.76, s = $350.90
Step 1: H0: = $1749, Ha: = $1749
Step 2: o = 0.05
Step 3:
1935.76 1749
2.661
350.9/ 25
t
= =
Step 4: Critical values = +2.064
Step 5: Since 2.661 > 2.064, reject H
0
. Note: For the P-value approach,
0.005 < P-value < 0.01. So, since the P-value < o , reject H
0
.
evidence to conclude that the mean annual expenditure on apparel
and services for consumer units in the Northeast differed from the
national mean in 2002.
AGE
F
r
e
q
u
e
n
c
y
100 90 80 70 60
5
4
3
2
1
0
Histogram of AGE
9.149 n = 187, df = 186, x
_
= 0.64, s = 0.15
Step 1: H0: = 0.9, Ha: < 0.9
Step 2: o = 0.05
Step 3: 703 . 23
187 / 15 . 0
90 . 0 64 . 0
=
= t
Step 5: Since 23.703 < -1.653, reject H
0
. Note: For the p-value
approach, P < 0.005. So, since the p-value < o , reject H
0
.
evidence to conclude that the mean ABI for women with
peripheral arterial disease is less than the healthy ABI of 0.9.
Thus, we conclude that such women do have an unhealthy ABI. The
practical significance of this result is that the ABI may be a
good tool for determining the possibility of peripheral arterial
disease in women. There could also be other causes of a low
ABI, so this test by itself may not be able to determine the
precise ailment.
9.150 n = 200, df = 199, x
_
= $2480, s = $766
Step 1: H0: = $2528, Ha: < $2528
Step 2: o = 0.05
Step 3:
2480 2528
0.886
766/ 200
t
= =
Step 5: Since 0.886 > -1.660, do not reject H
0
. Note: For the p-value
approach, P > 0.10. So, since the P-value < o , reject H
0
.
evidence to conclude that the mean cost of having a baby by AML
is less than the average cost of having a baby in a U.S.
hospital.
9.151 We used Minitab to produce the following histogram.
The sample size is only 20. Although there are no outliers, the
distribution is not very close to being normally distributed. It does not
appear to be reasonable to use a t-test with these data.
9.152 We used Minitab to produce the following normal probability plot.
BURIALS
P
e
r
c
e
n
t
3000 2000 1000 0 -1000 -2000
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
<0.005
292.8
S tDev 586.3
N 17
A D 3.364
P -V alu e
Probability Plotof BURIALS
Normal - 95% CI
SPENDING
P
e
r
c
e
n
t
20000 15000 10000 5000 0 -5000 -10000
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
<0.005
3462
S tDev 4182
N 27
A D 2.069
P -V alu e
Probability Plot of SPENDING
Normal - 95% CI
It is clear from the probability plot that the data are far from being
normally distributed. A t-test is clearly not appropriate for these data
with a sample size of only 17.
9.153 We used Minitab to produce the following normal probability plot.
9.154 It is clear from the probability plot that the data are far from being
normally distributed and the largest observation is a potential outlier. A
t-test is clearly not appropriate for these data with a sample size of only
27.
9.155 (a) Using Minitab, with the data in a column named PRESSURE, we choose Stat
Basic Statistics

1-Sample t..., click in the Samples in columns
text box and specify PRESSURE, click in the Test mean text box and
enter 80. Click the Options... button, enter 90 in the Confidence level
text box, click the arrow button at the right of the Alternative drop-
down list box and select greater than and click OK. Click on the
Graphs button and check the boxes for Histogram of Data and Boxplot of
data. Then click OK twice. Now choose Stat

Basic Statistics

Normality test and enter PRESSURE in the Variable text box. Click OK.
Then choose Graph

Stem-and-Leaf , enter PRESSURE in the Graph
variables text box and click OK. The results are
PRESSURE
P
e
r
c
e
n
t
120 110 100 90 80 70 60 50
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
0.625
81.76
StDev 10.32
N 41
AD 0.281
P-V alu e
Probability Plot of PRESSURE
Normal - 95% CI
PRESSURE
100 90 80 70 60
_
X
Ho
Boxplot of PRESSURE
(with Ho and 90% t-c onfi dence interval for the mean)
PRESSURE
F
r
e
q
u
e
n
c
y
100 90 80 70 60
8
6
4
2
0
_
X
Ho
Histogram of PRESSURE
(with Ho and 90% t-confidenc e interval for the mean)
Stem-and-leaf of PRESSURE N = 41
Leaf Unit = 1.0
1 5 8
2 6 3
5 6 569
10 7 00334
17 7 5677999
(8) 8 00111334
16 8 5899
12 9 0001334
5 9 5559
1 10 0
(b) The large sample size, lack of potential outliers, and the nearly
linear probability plot all indicate that a t-test is reasonable for
these data.
(c) The first procedure in part (a) also yielded the following test
results:
Test of mu = 80 vs > 80
90%
Lower
Variable N Mean StDev SE Mean Bound T P
PRESSURE 41 81.7561 10.3242 1.6124 79.6551 1.09 0.141
We see that t = 1.09 and the P-value is 0.141. Since the P-value is
greater than the significance level 0.10, we do not reject the null
hypothesis and conclude that the data do not provide evidence that the
mean diastolic blood pressure of bus drivers in Stockholm exceeds the
normal pressure of 88 mm Hg.
9.156 (a) Using Minitab, with the data in a column named DISTANCE, we choose Stat
Basic Statistics

text box and specify DISTANCE, click in the Test mean text box and type
80. Click the Options... button, enter 90 in the Confidence level text
box, click the arrow button at the right of the Alternative drop-down
list box and select greater than and click OK. Click on the Graphs
button and check the boxes for Histogram of Data and Boxplot of data.
Then click OK twice. Now choose Stat

Basic Statistics

Normality
DISTANCE
P
e
r
c
e
n
t
40 30 20 10 0 -10
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
<0.005
11.30
St Dev 6.002
N 500
A D 7.619
P- Value
Probability Plot of DISTANCE
Normal - 95% CI
DISTANCE
F
r
e
q
u
e
n
c
y
36 30 24 18 12 6
80
60
40
20
0
_
X
Ho
Histogramof DISTANCE
(with Ho and 95% t-c onfidence interval for the mean)
DISTANCE
40 30 20 10 0
_
X
Ho
Boxplot of DISTANCE
test and enter DISTANCE in the Variable text box. Click OK. Then
choose Graph

Stem-and-Leaf , enter DISTANCE in the Graph variables
text box and click OK. The results are
Stem-and-leaf of DISTANCE N = 500
Leaf Unit = 1.0
4 0 1111
27 0 22222222333333333333333
92 0 44444444444444444444444444444455555555555555555555555555555555555+
169 0 66666666666666666666666666666666666777777777777777777777777777777+
248 0 88888888888888888888888888888888888888889999999999999999999999999+
(58) 1 0000000000000000000000000000000000111111111111111111111111
194 1 22222222222222222222222222223333333333333333333333333333333
135 1 4444444444444445555555555555555555
101 1 6666666666666666777777777777777777
67 1 8888888888999999999
48 2 0000000000011111111
29 2 22223333
21 2 44444455
13 2 6
HI 26, 26, 27, 27, 27, 28, 28, 29, 29, 30, 35, 36
(b) The large sample size makes this set of data a reasonable candidate for
a t-test. However, the large number of outliers (12) indicate that
some caution is appropriate. If there is any doubt, remove the outliers
and retest.
(c) The first procedure in part (a) also yielded the following test
results:
Variable N Mean StDev SE Mean 95% CI T P
DISTANCE 500 11.3020 6.0023 0.2684 (10.7746, 11.8294) -2.23 0.026
The P-value of the test is 0.026, which is less than the significance
level of 0.05. We reject the null hypothesis and conclude that the
data do provide sufficient evidence that the mean distance driven last
year differs from that in 2000. Note: If the 12 large outliers are
all removed, the t-value changes to 4.34 with a P-value of 0.000, so
our conclusion does not change.
9.157 (a) Using Minitab, with the data in a column named DISTANCE, we choose Stat
Basic Statistics

text box and specify RENT, click in the Test mean text box and enter
692. Click the Options... button, enter 95 in the Confidence level
text box, click the arrow button at the right of the Alternative drop-
down list box and select greater than and click OK.
95%
Lower
RENT 100 703.960 89.846 8.985 689.042 1.33 0.093
The P-value for the test is 0.093, larger than the significance lever
0.05, so we do not reject the null hypothesis. The data do not provide
sufficient evidence to conclude that the mean rent for a two-bedroom
unit in Maine is greater than the FMR of $692.
(b) After removing the outlier 405 and following the procedure in part (a),
the results are
95%
Lower
RENT 99 706.980 85.049 8.548 692.786 1.75 0.041
(c) Now the P-value is 0.041, leading to rejection of the null hypothesis.
The sample mean increased by $3.02 and the standard deviation decreased
by about 4.8.
(d) The sample size is large in both cases, yet the effect of the outlier
is considerable. Caution should be used and perhaps the data should be
analyzed using a method that is not influenced by outlers.
9.158 (a) Neither the z-test nor the t-test is appropriate for small samples
containing outliers or exhibiting extremely non-normal distributional
shapes.
(b) A non-parametric test that is neither affected by a few outliers nor by
non-normality might be appropriate.
9.159 (a) n = 20, df = 19, t
/2
= 1.729, s = 2.291, x
_
= 4.835
The hypothesized mean (4.66) lies within the confidence interval, so we
should not reject the null hypothesis. Since the test statistic is
(4.835 4.66)/(2.291/ 20) 0.342 t = = , which is less than the critical value
(b) n = 25, df = 24, t
/2
= 2.064, s = $350.90, x
_
= $1935.76
The 95% confidence interval is
1935.76 2.064(350.90)/ 25 (1790.91,2080.61) = .
The hypothesized mean ($1749) lies outside the confidence interval, so
(1935.76 1749)/(350.90/ 25) 2.661 t = = , which is greater than the critical
value of 2.064, the hypothesis test also leads to the conclusion that
we should reject the null hypothesis.
9.160 (a) n = 187, df = 186, -t
= -1.660, s = 0.15, x
_
= 0.64
The 95% upper confidence bound is 0.64 1.660(0.15)/ 187 0.658 + = .
The hypothesized mean (0.9) lies above the upper confidence bound, so
(0.64 0.9)/(0.15/ 187) 23.703 t = = , which is less than the critical value
(b) n = 200, df = 199, -t
= -1.660, s = $766, x
_
= $2480
The 95% upper confidence bound is 2480 1.660(766)/ 200 2569.91 + = .
The hypothesized mean ($2528) lies below the upper confidence bound, so
we not should reject the null hypothesis. Since the test statistic is
(2480 2528)/(766/ 200) 0.886 t = = , which is greater than the critical
value of 1.660, the hypothesis test also leads to the conclusion that
we should not reject the null hypothesis.
9.161 (a) n = 6, df = 5, t
= 2.015, s = 2.7, x
_
= 182.7
The 95% lower confidence bound is 182.7 2.015(2.7)/ 6 180.479 = .
The hypothesized mean (180) lies below the lower confidence bound, so
(182.7 180)/(2.7/ 6) 2.449 t = = , which is greater than the critical value
The hypothesized mean (180) lies above the lower confidence bound, so
(182.7 180)/(2.7/ 6) 2.449 t = = , which is less than the critical value of
not reject the null hypothesis.
(b) n = 10, df = 9, t
= 2.821, s = 0.149, x
_
= 2.5
Section 9.7, The Wilcoxon Signed-Rank Test 429
The hypothesized mean (2.30) lies below the lower confidence bound, so
(2.5 2.3)/(0.149/ 10) 4.245 t = = , which is greater than the critical value
Exercises 9.7
9.162 Technically, nonparametric methods are inferential methods that are not
concerned with parameters. In practice, nonparametric methods are those
that can be applied without assuming normality.
9.163 The advantages of nonparametric methods are that they do not require
normality, they make use of fewer and simpler calculations than do
parametric methods, and they are resistant to outliers. The disadvantage of
nonparametric methods is that they tend to give less accurate results than
parametric methods when the assumptions underlying the parametric methods
are actually met.
9.164 The population must be symmetric.
9.165 Because the D-value for such a data value equals 0, we cannot attach a sign
to the rank of |D|.
9.166 (a) Wilcoxon signed-rank test (b) t-test
(c) Neither
9.167 (a) Wilcoxon signed-rank test (b) Wilcoxon signed-rank test
(c) Neither
9.168 It is because the median and the mean are the same when considering a
symmetric distribution.
9.169 (a) W
0.05
= 30 (b) W
0.95
= 8(8 + 1)/2 30 = 6
(c) W0.025 = 32 W0.975 = 8(8 + 1)/2 32 = 4
9.170 (a) W0.01 = 50 (b) W0.95 = 10(10 + 1)/2 50 = 5
(c) W0.005 = 52 W0.995 = 10(10 + 1)/2 52 = 3
9.171 (a) W0.10 = 128 (b) W0.90 = 19(19 + 1)/2 128 = 62
(c) W0.05 = 136 W0.995 = 19(19 + 1)/2 136 = 54
9.172 (a) W0.05 = 90 (b) W0.95 = 15(15 + 1)/2 90 = 30
(c) W0.025 = 95 W0.975 = 15(15 + 1)/2 95 = 25
9.173 H0: = 124.9 days, Ha: < 124.9 days; n = 8; = 0.05,
W0.95 = 8(8 + 1)/2 30 = 6
Days D |D| Rank |D| R
103 -21.9 21.9 4 -4
80 -44.9 44.9 5 -5
79 -45.9 45.9 7 -7
135 10.1 10.1 2 2
134 9.1 9.1 1 1
77 -47.9 47.9 8 -8
80 -44.9 44.9 5 -5
111 -13.9 13.9 3 -3
The sum of the positive ranks is W = 3. This is less than the critical
value 6, so we reject the null hypothesis and conclude that the data provide
sufficient evidence that the mean number of days of ice cover on Lake Wingra
is less than it was in the late 1800s.
9.174 H
0
: = 46.7 years, H
a
: =/46.7 years; n = 8; = 0.05, W
0.025
= 32,
W
0.975
= 8(8 + 1)/2 32 = 4
Years D |D| Rank |D| R
30.3 -16.4 16.4 7 -7
47.0 0.3 0.3 1 1
56.4 9.7 9.7 5 5
30.5 -16.2 16.2 6 -6
39.6 -7.1 7.1 4 -4
47.9 1.2 1.2 2 2
29.7 -17 17.0 8 -8
52.5 5.8 5.8 3 3
The sum of the positive ranks is W = 11. This is between the critical
values 4 and 32, so we do not reject the null hypothesis and conclude that
the data do not provide sufficient evidence that the mean number of happy
life years has changed from that in the 1900s.
9.175
0
q = 35.7, n = 10, o = 0.01
Step 1: H
0
: q = 35.7, H
a
: q > 35.7
Step 2: o = 0.01
Step 3:
x
x-
0
q =D
|D| Rank of |D| Signed Rank R
42 6.3 6.3 3 3
45 9.3 9.3 4 4
62 26.3 26.3 10 10
49 13.3 13.3 6 6
14 -21.7 21.7 8 -8
39 3.3 3.3 2 2
57 21.3 21.3 7 7
11 -24.7 24.7 9 -9
36 0.3 0.3 1 1
26 -9.7 9.7 5 -5
Step 4: W = sum of the + ranks = 33
Step 5: Critical value = 50
Step 6: Since W < 50, do not reject H0.
evidence to conclude that the median age has increased over the
2002 median age of 35.7 years. The P-value is 0.305.
9.176
0
= 254, n = 12, o = 0.05
Step 1: H
0
: = 250, H
a
: > 250
Step 2: o = 0.05
Step 3:
x
x-
0
=D
395 141 141 12 12
274 20 20 1 1
210 -44 44 8 -8
307 53 53 10 10
218 -36 36 4 -4
293 39 39 6.5 6.5
299 45 45 9 9
283 29 29 3 3
293 39 39 6.5 6.5
228 -26 26 2 -2
315 61 61 11 11
292 38 38 5 5
Step 6: Since W > 61, reject H0.
evidence to conclude that last year's mean amount spent by
consumers on nonalcoholic beverages has increased over the 2002
mean of $250. The P-value is 0.027.
9.177
0
= $12,850, n = 10, o = 0.10
Step 1: H
0
: = $12,850, H
a
: < $12,850
Step 2: o = 0.10
Step 3:
Price D |D|Rank |D| R
13480 630 630 8 8
12992 142 142 3 3
12988 138 138 2 2
12800 -50 50 1 -1
12599 -251 251 5 -5
12499 -351 351 7 -7
11500 -1350 1350 9 -9
10400 -2450 2450 10 -10
12500 -350 350 6 -6
12600 -250 250 4 -4
Step 5: Critical value = 10(11)/2 - 41 = 14
Step 6: Since W < 14, reject H
0.
evidence to conclude that the mean asking price for a 2003 Ford
Mustang is less than the 2006 Kelly Blue Book value. The P-value
is 0.077.
9.178
0
q = 7.4, n = 13, o = 0.05 (The value of 7.4 was deleted from the original
sample.)
Step 1: H0: q = 7.4, Ha: q = 7.4
Step 2: o = 0.05
Step 3:
x
x-
0
q =D
D Rank of D Signed Rank R
8.6 1.2 1.2 4 4
8.8 1.4 1.4 5.5 5.5
8.2 0.8 0.8 3 3
5.3 -2.1 2.1 11 -11
9.2 1.8 1.8 9 9
13.8 6.4 6.4 13 13
5.6 -1.8 1.8 9 -9
7.8 0.4 0.4 2 2
6.0 -1.4 1.4 5.5 -5.5
5.7 -1.7 1.7 7 -7
11.6 4.2 4.2 12 12
9.2 1.8 1.8 9 9
7.2 -0.2 0.2 1 -1
Step 4: W = sum of the + ranks = 57.5
Step 5: Critical values = 17, 74 [13(14)/2 - 74 = 17]
Step 6: Since 17 < W < 74, do not reject H
0
.
evidence to conclude that this year's median birth weight differs
from the median birth weight of 7.4 in 2002. The P-value is 0.422.
9.179
0
= 2.30, n = 10, o = 0.01
(a) Step 1: H
0
: = 2.30, H
a
: > 2.30
Step 2: o = 0.01
Step 3:
x x-
0
=D |D| Rank of |D| Signed Rank R
2.41 0.11 0.11 3 -3
2.60 0.30 0.30 8 8
2.31 0.01 0.01 1 1
2.51 0.21 0.21 5.5 5.5
2.54 0.24 0.24 7 7
2.51 0.21 0.21 5.5 5.5
2.28 -0.02 0.02 2 -2
2.42 0.12 0.12 4 4
2.72 0.42 0.42 10 10
2.70 0.40 0.40 9 9
Step 6: Since W > 50, reject H
0
.
evidence to conclude that the mean available limestone in
soil treated with 100% MMBL effluent exceeds 2.30%. The
P-value is 0.005.
(b) A Wilcoxon signed-rank test is permissible because a normally
distributed population is symmetric.
9.180
0
= 2, n = 13, o = 0.05 (All values of 2 were deleted from the data.)
(a) Step 1: H0: = 2, Ha: > 2
Step 2: o = 0.05
Step 3:
x
x-
0
=D
D Rank of D Signed Rank R
3 1 1 6.5 6.5
3 1 1 6.5 6.5
3 1 1 6.5 6.5
1 -1 1 6.5 -6.5
3 1 1 6.5 6.5
3 1 1 6.5 6.5
3 1 1 6.5 6.5
4 2 2 13.0 13.0
3 1 1 6.5 6.5
1 -1 1 6.5 -6.5
3 1 1 6.5 6.5
3 1 1 6.5 6.5
1 -1 1 6.5 -6.5
Step 6: Since W > 70, reject H
0
.
evidence to conclude that Helsinkians do respond with an
ecological welfare food choice motive greater than 2. The P-
value is 0.037.
(b) For the t test, we have n = 18, df = 17, x
_
= 44/18 = 2.4444, and s =
0.85559.
o = 0.05 and the critical value is 1.740.
t
x
s n
=
0
2 4444 2
0 85559 18
2 204
/
.
. /
.
Since 2.204 > 1.740, we reject the null hypothesis. There is
sufficient evidence to conclude that Helsinkians respond with an
ecological welfare food choice motive greater than 2.
(c) The two tests lead to the same conclusion. However, since the data
were limited to integer values between 1 and 4, it is not possible for
this data to have come from a normal distribution, or even an
approximately normal distribution. The Wilcoxon test is a better
choice for a test with these data.
9.181 n = 16, x
_
= 306, s = 8.671, o = 0.05
(a) Step 1: H0: = 310, Ha: < 310
Step 2: o = 0.05
Step 3: 845 . 1
16 / 671 . 8
310 306
=
= t
Step 5: Since -1.845 < -1.753, reject H0. Note: For the P-value
approach, 0.025 < p-value < 0.05. So, since P-value < o ,
reject H0.
evidence to conclude that the mean content, , is less than
the advertised content of 310 ml.
(b) Step 1: H
0
: = 310, H
a
: < 310
Step 2: o = 0.05
Step 3:
x
x-
0
=D
297 -13 13 14 -14
311 1 1 2 2
322 12 12 12.5 12.5
315 5 5 7 7
318 8 8 9 9
303 -7 7 8 -8
307 -3 3 5 -5
296 -14 14 15 -15
306 -4 4 6 -6
291 -19 19 16 -16
312 2 2 4 4
309 -1 1 2 -2
300 -10 10 10.5 -10.5
298 -12 12 12.5 -12.5
300 -10 10 10.5 -10.5
311 1 1 2 2
Step 6: Since W > 36, do not reject H
0
.
Step 7: At the 5% significance level, the data do not provide
sufficient evidence to conclude that the mean content, , is
less than the advertised content of 310 ml. The P-value is
0.054.
(c) Since the population is normally distributed, the t-test is more
powerful than the Wilcoxon signed-rank test; that is, the t-test is
more likely to detect a false null hypothesis.
9.182 n = 10, x
_
= 8.9, s = 2.4698, o = 0.10,
0
q = 10.2
(a) Step 1: H0: q = 10.2, Ha: q = 10.2
Step 2: o = 0.10
Step 3: 664 . 1
10 / 4698 . 2
2 . 10 9 . 8
=
= t
Step 5: Since -1.833 < -1.664 < 1.833, do not reject H
0
. Note: For
the P-value approach, 0.10 < p-value < 0.20. So, since P-
value > o , do not reject H
0
.
sufficient evidence to conclude that this year's median
educational attainment has changed from the median 25 years
ago.
(b) Step 1: H0: q = 10.2, Ha: q = 10.2
Step 2: o = 0.10
Step 3:
x
x-
0
q =D
14 3.8 3.8 8 8
10 -0.2 0.2 2 -2
10 -0.2 0.2 2 -2
10 -0.2 0.2 2 -2
5 -5.2 5.2 10 -10
8 -2.2 2.2 6.5 -6.5
6 -4.2 4.2 9 -9
9 -1.2 1.2 4.5 -4.5
8 -2.2 2.2 6.5 -6.5
9 -1.2 1.2 4.5 -4.5
Step 5: Critical values = 11, 44
Step 6: Since W < 11, reject H
0
.
evidence to conclude that this year's median educational
attainment has changed from the median 25 years ago of 10.2
years.
(c) Since the population is given to be symmetric and nonnormal, the
Wilcoxon signed-rank test is more powerful than the t-test and thus is
more likely to detect a false null hypothesis.
9.183 (a) Using Minitab, with the data in a column named SCORES, we choose Stat

Nonparametrics 1-Sample Wilcoxon..., select SCORES in the Variables
text box, click in the Test median box and type 5, click the arrow
button at the right of the Alternative drop down list box, and select
Less than, and click OK. The result is
Test of median = 5.000 versus median < 5.000
N
for Wilcoxon Estimated
N Test Statistic P Median
SCORE 156 84 777.0 0.000 4.500
Since the P-value of 0.000 is less than the significance level of 0.01,
we reject the null hypothesis and conclude that there is very strong
evidence that professional golfers score better than par on the Hole
OCross Out.
(b) To perform the t-test with the original data, we choose Stat

Basic
statistics

1-Sample t..., select the Samples in columns box, enter
SCORES in the Samples in Columns text box, enter 5 in the Test Mean
text box, click on the Options button, enter 99.0 in the Confidence
level text box, select less than in the Alternative box, and click OK
twice. The results in the Session Window are
99%
Upper
SCORES 156 4.66667 0.76482 0.06123 4.81061 -5.44 0.000
Again, since the P-value of 0.000 is less than the significance level
of 0.01, we reject the null hypothesis and conclude that there is very
strong evidence that professional golfers score better than par on the
Hole OCross Out.
(c) The results are the same for the two tests. The Wilcoxon test is the
more appropriate test, however, since the data are discrete and cannot
have a normal distribution.
9.184 Using Minitab, with the data in a column named PRESSURE, we choose Stat

Nonparametrics

1-Sample Wilcoxon..., select PRESSURE in the Variables
text box, click in the Test median box and type 80, click the arrow button
at the right of the Alternative drop down list box, and select greater than,
and click OK. The result is
Test of median = 80.00 versus median > 80.00
N
PRESSURE 41 39 468.0 0.140 82.00
Since the P-value of 0.140 is greater than the significance level of 0.10,
we do not reject the null hypothesis and conclude that there is weak or no
evidence that the mean diastolic blood pressure of bus drivers in Stockholm
exceeds the normal diastolic blood pressure of 80 mm Hg.
9.185 Using Minitab, with the data in a column named DISTANCE, we choose Stat

Nonparametrics

1-Sample Wilcoxon..., select DISTANCE in the Variables
text box, click in the Test median box and type 11.9, click the arrow button
at the right of the Alternative drop down list box, and select not equal,
Test of median = 11.90 versus median not = 11.90
N
DISTANCE 500 497 49193.0 0.000 10.75
Since the P-value of 0.000 is less than the significance level 0.05, we
reject the null hypothesis and conclude that the mean miles driven by cars
last year differs from the mean distance driven in 2000.
9.186 Using Minitab, with the data in a column named RENT, we choose Stat

Nonparametrics

1-Sample Wilcoxon..., select RENT in the Variables text
box, click in the Test median box and type 692, click the arrow button at
the right of the Alternative drop down list box, and select greater than,
Test of median = 692.0 versus median > 692.0
N
RENT 100 99 2873.5 0.082 705.0
Since the P-value of 0.082 is greater than the significance level of 0.05,
we do not reject the null hypothesis and conclude that the data do not
provide sufficient evidence that the mean monthly rent for two bedroom units
in Maine is greater than the FMR of $692.
9.187 The Wilcoxon signed-rank test is likely to give better results. The
distribution of marriage durations is unlikely to be normal, possibly not
even symmetric. Given that duration cannot be less than 0 years and there
are likely to be some fairly long marriages which might look like outliers,
it would be better to use the Wilcoxon signed-rank test which is insensitive
to outliers than to use the t-test which assumes normality and is sensitive
to outliers.
9.188 The distribution of times to complete the U.S. Census Form may be
approximately normal, but it is also likely that there will be some people
who unfamiliar with the form, dont know the answers to some questions, have
poor eyesight, or are very slow or poor readers, but one would not expect
that there would be many extremely long times recorded. An examination of
the data might prove helpful in evaluating the symmetry of the population.
Given the uncertainty of the normality assumption for the t-test and the
potential for outliers, it would be better to use the Wilcoxon signed-rank
test.
9.189 (a) If John is not unlucky, he should expect to wait 15 minutes for the
train, on the average.
(b) If John is not unlucky, the distribution of the times he waits for the
trains should be a uniform distribution over the interval from 0 to 30
minutes.
(c) Step 1: H0: q = 15, Ha: q > 15
Step 2: o = 0.10
Step 4:
x
x-
0
q =D
D Rank of D
Signed Rank
R
24 9 9 6.5 6.5
26 11 11 9.5 8.5
20 5 5 3.5 3.5
4 -11 11 9.5 -9.5
20 5 5 3.5 3.5
3 -12 12 11 -11
19 4 4 2 4
5 -10 10 8 -8
28 13 13 12 12
16 1 1 1 1
22 7 7 5 5
24 9 9 6.5 6.5
Step 6: Since W = 50.5 < 56, do not reject H
0
.
sufficient evidence to conclude that John waits more than 15
minutes for the train, on the average.
(d) Since the population is uniform (which is symmetric), the Wilcoxon test
is appropriate.
(e) Since the population is symmetric and nonnormal, the Wilcoxon signed-
rank test is more powerful than the t-test and more appropriate than
the t-test, which assumes normality.
9.190 Step 1: State the null and alternative hypotheses.
Step 2: Decide on the significance level o .
Step 3: Construct a worktable of the form:
Data
Value
X
Difference
D = x -
0
D
Rank
of D
Signed
Rank
R
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Step 4: Compute the value of the test statistic:
( 1)/ 4
( 1)(2 1)/ 24
W n n
z
n n n
+
=
+ +
where W is the sum of the positive ranks.
Step 5: The critical value(s):
(a) for a two-tailed test are
/2
z
o
.
(b) for a left-tailed test is - z
o
(c) for a right-tailed test is z
o
.
Use Table II to find the critical value(s).
Step 6: If the value of the test statistic falls in the rejection region,
reject H
0
; otherwise, do not reject H
0
.
Step 7: State the conclusion in words.
9.191 (a) Step 1: H0: q = 7.4 lb, Ha: q = 7.4 lb
Step 2: o = 0.05
Step 3: See Step 3 in the solution to Exercise 9.178.
Step 4: From Step 4 in the solution to Exercise 9.178, W = 57.5. Now:
84 . 0
24 / ) 1 13 2 )( 1 13 ( 13
4 / ) 1 13 ( 13 5 . 57
24 / ) 1 2 )( 1 (
4 / ) 1 (
=
+ +
+
=
+ +
+
=
n n n
n n W
z
Step 6: Since -1.96 < 0.84 < 1.96, do not reject H0.
evidence to conclude that this year's median birth weight
differs from that in 2000.
6 5 4 3 2 1 0
0. 25
0. 20
0. 15
0. 10
0. 05
0. 00
W
F
R
E
Q
(b) Neither the Wilcoxon signed-rank test nor the normal approximation led
9.192 (a) Summing the ranks corresponding to the "+" signs in each row results in
a value for W. All possible values for W are presented in the last
column.
Rank
1 2 3 W
+ + + 6
+ + - 3
+ - + 4
+ - - 1
- + + 5
- + - 2
- - + 3
- - - 0
(b) Since there are eight equally likely outcomes, the probability that a
sample will match any particular row of the table is 1/8 = 0.125.
(c) The probability distribution (d) A histogram for the probability
of the random variable W distribution of W for n = 3 is
when n = 3 is:
(e) For left-tailed tests with n = 3 and = 0.125, the critical value
W = 0.
9.193 Summing the ranks corresponding to the "+" signs in each row results in a
value for W. All sixteen possible values for W are presented in the last
column.
W P(W)
0 0.125
1 0.125
2 0.125
3 0.250
4 0.125
5 0.125
6 0.125
10 9 8 7 6 5 4 3 2 1 0
0.10
0.05
0.00
W
F
R
E
Q
(a) (b) 1/16 = 0.0625
Rank (c)
1 2 3 4 W W P(W)
+ + + + 10 0 0.0625
+ + + - 6 1 0.0625
+ + - + 7 2 0.0625
+ + - - 3 3 0.1250
+ - + + 8 4 0.1250
+ - + - 4 5 0.1250
+ - - + 5 6 0.1250
+ - - - 1 7 0.1250
- + + + 9 8 0.0625
- + + - 5 9 0.0625
- + - + 6 10 0.0625
- + - - 2
- - + + 7
- - + - 3
- - - + 4
- - - - 0
(d)
(e) For a left-tailed test with n = 4 and o = 0.125, the critical value W
equals 1 since W = 0 or 1 has probability 0.125.

9.194 (a) 0 is the hypothesized median. The median is that value which has half
of the population to its left and half to its right. Therefore, the
probability that a value exceeds the median is 0.5.
(b) Each observation either exceeds 0 or it doesnt. The observations are
independent of each other and the probability p that an observation
exceeds 0 is 0.5 for each observation. The number of observations
(trials) is fixed at n. Thus, the number of observations exceeding 0
satisfies all of the criteria for a binomial distribution.
9.195 (a) The Wilcoxon signed-rank test takes into account the sign and the
absolute size of each difference from the median, whereas, the sign
test only considers the sign. Since the Wilcoxon test uses more
information than the sign test, it is more likely to be able to detect
a false null hypothesis.
(b) The sign test can be used with any distribution since the probability
that an observation exceeds the median is always 0.5, regardless of the
shape of the distribution. The Wilcoxon signed-rank test is based on
an assumption that the underlying distribution of the data is
symmetric, a slightly more restrictive assumption than for the sign
test.
9.196 If one or more observations equals
0
, delete those observations and
continue with the testing process, using a reduced sample size.
9.197 (a) For the sign test, x = 2 of the n = 8 observations are above 124.9.
The P-value is P(x < 2) for a binomial distribution with n = 8 and p =
0.5. From Table XII, this probability is 0.004 + 0.031 + 0.109 =
0.144. Since this is greater than the 0.05 significance level, the
data do not provide sufficient evidence that the number of days the
lake was frozen over is less now than in the late 1800s.
(b) The Wilcoxon signed rank test had W = 3.0 with a P-value of 0.021.
Since this is less than the 0.05 significance level, the data do
provide sufficient evidence that the number of days the lake was frozen
over is less now than in the late 1800s.
9.198 (a) For the sign test, x = 4 of the n = 8 observations are above 46.7. The
P-value is 2min{P(x < 4), P(x < 4)} for a binomial distribution with n
= 8 and p = 0.5. From Table XII, both probabilities in the {} are
0.636. Multiplying by 2, we get a probability greater than 1, which we
just adjust to 1.000. Since this is greater than the 0.05 significance
level, the data do not provide sufficient evidence that the median
number of happy-life years has changed from that in the 1900s.
Since this is greater than the 0.05 significance level, the data do not
provide sufficient evidence that the median number of happy-life years
has changed from that in the 1900s. Same conclusion, smaller P-value.
9.199 (a) For the sign test, x = 7 of the n = 10 observations are above 35.7.
The P-value is P(x > 7)} for a binomial distribution with n = 10 and p
= 0.5. From Table XII, this probability is 0.117 + 0.044 + 0.010 +
0.001 = 0.172. Since this is greater than the 0.05 significance level,
the data do not provide sufficient evidence that the median age of
todays U.S. residents has increased from the 2002 median of 35.7
years.
provide sufficient evidence that the median age of todays U.S.
residents has increased from the 2002 median of 35.7 years. Same
conclusion, larger P-value.
9.200 (a) For the sign test, x = 9 of the n = 12 observations are above 254. The
P-value is P(x > 9)} for a binomial distribution with n = 12 and p =
0.5. From Table XII, this probability is 0.054 + 0.016 + 0.003+ 0.000
= 0.073. Since this is greater than the 0.05 significance level, the
data do not provide sufficient evidence that last years mean amount
spent by consumers on nonalcoholic beverages has increased from the
2002 mean of $254.
provide sufficient evidence that last years mean amount spent by
consumers on nonalcoholic beverages has increased from the 2002 mean of
$254. Different conclusion, smaller P-value.
9.201 (a) For the sign test, x = 3 of the n = 10 observations are above $12850.
The P-value is P(x < 3)} for a binomial distribution with n = 10 and p
= 0.5. From Table XII, this probability is 0.001 + 0.044 + 0.010 +
0.001 = 0.172. Since this is greater than the 0.10 significance level,
the data do not provide sufficient evidence that the mean asking price
for 2003 Ford Mustang coupes in Phoenix is less than the 2006 Kelley
Blue Book retail value of $12850.
provide sufficient evidence that the mean asking price for 2003 Ford
Mustang coupes in Phoenix is less than the 2006 Kelley Blue Book retail
value of $12850. Different conclusion, smaller P-value.
9.202 (a) For the sign test, x = 8 of the n = 13 observations (the observation
7.4 was deleted) are above 7.4 lbs. The P-value is 2min{P(x < 8), P(X
> 8)}} for a binomial distribution with n = 13 and p = 0.5. From Table
XII, P(X < 80) =.866 and P(X > 8) = 0.291. Therefore, P = 2(0.291) =
0.582. Since this is greater than the 0.05 significance level, the
data do not provide sufficient evidence that this years median birth
weight differs from the 2002 median birth weight.
provide sufficient evidence that this years median birth weight
differs from the 2002 median birth weight. Same conclusion, smaller P-
value.
Exercises 9.8
9.203 (a) One-mean z-test, One-mean t-test, Wilcoxon signed-rank test
(b) The z-test assumes that o is known, and that the population is normal
or the sample is large. The t-test assumes that o is unknown, and
that the population is normal or the sample is large. The signed-rank
test assumes only that the population is symmetric.
(c)
0 0
( )/( / ) t ( )/( / ) z x n x s n o = =
W = the sum of the positive ranks
9.204 (a) Yes. The t-test can be used when the population is normal and the
population standard deviation is unknown.
(b) Yes. The normal distribution is symmetric and that is all that is
required in order to use the Wilcoxon signed-rank test.
(c) The t-test is preferable in this situation since it is more powerful
(more likely to detect a false null hypothesis) when the population is
normal.
9.205 (a) Yes. The t-test can be used when the sample size is large. It is
almost equivalent to the z-test in this situation.
(b) Yes. The Wilcoxon signed-rank test can be used when the population
distribution is symmetric.
(c) The Wilcoxon signed-rank test is preferable in this situation since it
is more powerful (more likely to detect a false null hypothesis) when
the population is symmetric, but nonnormal.
9.206 (a) Yes. It is permissible to use the z-test when the sample size is large,
Chapter 9 Review Problems 443
even if the population is skewed.
(b) No. The Wilcoxon signed-rank test is valid when the population
distribution is symmetric, but not when it is highly skewed.
9.207 Since we have normality and o is known, use the z-test.
9.208 Since we have a large sample, use the z-test.
9.209 Since we have a large sample with no outliers and o is unknown, use the t-
test.
9.210 Since we have a symmetric non-normal distribution, use the Wilcoxon signed-
rank test.
9.211 Since we have a symmetric non-normal distribution, use the Wilcoxon signed-
rank test.
9.212 Since we have a symmetric distribution, use the Wilcoxon signed-rank test.
9.213 The distribution looks skewed and the sample size is not large. Consult a
statistician.
9.214 The distribution looks skewed and the sample size is not large. Consult a
statistician.
Review Problems for Chapter 9
1. (a) A null hypothesis always specifies a single value for the parameter of
a population which is of interest.
(b) The alternative hypothesis reflects the purpose of the hypothesis test,
which can be to determine that the parameter of interest is greater
than, less than, or different from the single value specified in the
null hypothesis.
(c) The test statistic is a quantity calculated from the sample, under the
assumption that the null hypothesis is true, which is used as a basis
for deciding whether or not to reject the null hypothesis.
(d) The rejection region is a set of values of the test statistic that lead
(e) The nonrejection region is a set of values of the test statistic that
lead to not rejecting the null hypothesis.
(f) The critical values are values of the test statistic that separate the
rejection region from the nonrejection region.
2. (a) The statement is expressing the fact that there is variability in the
net weights of the boxes content and some boxes may actually contain
less than the printed weight on the box. However, the net weights for
each days production will average a bit more than the printed weight.
(b) To test the truth of this statement, we would use a null hypothesis
that stated that the population mean net weight of the boxes was equal
to the printed weight and an alternative hypothesis that stated that
the population mean net weight of the boxes was greater than the
printed weight.
(c) Null hypothesis: Population mean net weight = 76 oz
Alternative hypothesis: Population mean net weight > 76 oz
or
H
0
: = 76 H
a
: > 76
3. (a) Roughly speaking, there is a range of values of the test statistic
which one could reasonably expect to occur if the null hypothesis were
true. If the value of the test statistic is one that would not be
expected to occur when the null hypothesis is true, then we reject the
null hypothesis.
(b) To make this procedure objective and precise, we specify the
probability with which we are willing to reject the null hypothesis
when it is actually true. This is called the significance level of the
test and is usually some small number like 0.05 or 0.01. Specifying
the significance level allows us to determine the range of values of
the test statistic that will lead to rejection of the null hypothesis.
If the computed value of the test statistic falls in this rejection
region, then the null hypothesis is rejected. If it does not fall in
the rejection region, then the null hypothesis is not rejected.
4. We would use the alternative hypothesis =
0
if we wanted to determine
whether the population mean were different from the value 0 specified in
the null hypothesis. We would use the alternative hypothesis >
0
if we
wanted to determine whether the population mean were greater than from the
value
0
specified in the null hypothesis. We would use the alternative
hypothesis <
0
if we wanted to determine whether the population mean
were less than the value 0 specified in the null hypothesis.
5. (a) A Type I error is made whenever the null hypothesis is true, but the
value of the test statistic leads us to reject the null hypothesis. A
Type II error is made whenever the null hypothesis is false, but the
value of the test statistic leads us to not reject the null hypothesis.
(b) The probability of a Type I error is represented by o and that of a
Type II error by | .
(c) If the null hypothesis is true, the test statistic can lead us to
either reject or not reject the null hypothesis. The first is the
correct decision, while the latter constitutes a Type I error. Thus a
Type I error is the only type of error possible when the null
hypothesis is true.
(d) If the null hypothesis is not rejected, a correct decision has been
made if the null hypothesis is, in fact, true. But if the null
hypothesis is false, we have made a Type II error. Thus a Type II
error is the only type of error possible when the null hypothesis is
not rejected.
6. Assuming that the null hypothesis is true, find the value of the test
statistic for which the probability of obtaining a value greater than the
specified value is 0.05.
7. (a) If the population standard deviation is unknown, and the population is
normal or the sample size is large, we can use the one-mean t-
statistic, ) / /( ) (
0
n s x t = .
(b) If the population standard deviation is known, and the population is
normal or the sample size is large, we can use the one-mean z-
statistic, ) / /( ) (
0
n x z o = .
(c) If the population is symmetric, we can use the Wilcoxon signed-rank
statistic W = the sum of the positive ranks.
8. (a) A hypothesis test is exact if the actual significance level is the same
as the one that is stated.
(b) A hypothesis test is approximately correct if the actual significance
level only approximately equals o .
9. A statistically significant result occurs when the value of the test
statistic falls in the rejection region. A result has practical
significance when it is statistically significant and the result also is
different enough from results expected under the null hypothesis to be
important to the consumer of the results. By taking large enough sample
sizes, almost any result can be made statistically significant due to the
increased ability of the test to detect a false null hypothesis, but small
differences from the conditions expressed by the null hypothesis may not be
important, that is, they may not have practical significance.
10. The probability of a Type II error is increased when the significance level
is decreased for a fixed sample size.
11. (a) The power of a hypothesis test is the probability of rejecting the null
hypothesis when the null hypothesis is false.
(b) The power of a test increases when the sample size is increased while
keeping the significance level constant.
12. (a) The P-value of a hypothesis test is the probability, assuming that the
null hypothesis is true, of getting a value of the test statistic that
is as extreme or more extreme than the one actually obtained.
(b) True. If the null hypothesis were true, a value of the test statistic
with a P-value of 0.02 would be more extreme than one with a P-value of
0.03.
(c) True. If the P-value is 0.74, this means that 74% of the time when the
null hypothesis is true, the value of the test statistic would be more
extreme than the one actually obtained.
(d) The P-value of a hypothesis test is also called the observed
significance level since it represents the smallest possible
significance level at which the null hypothesis could have been
rejected.
13. In the critical-value approach, the null hypothesis is rejected if the value
of the test statistic falls in the rejection region that is determined by
the chosen significance level. In the P-value approach, the test statistic
is computed and then the probability of obtaining a value as extreme or more
extreme than the one actually obtained is found. This is the P-value. The
advantages of providing the P-value are that the observed significance level
of the of the test is given and the reader of the results can determine for
him/herself whether the results are strong enough evidence against the null
hypothesis to reject it.
14. Non-parametric methods have the advantages of involving fewer and simpler
calculations than parametric methods and are more resistant to outliers and
other extreme values. Parametric methods are preferred when the population
is normal or the sample size is large since they are more powerful than non-
parametric methods and thus tend to give more accurate results than non-
parametric methods under those conditions.
15. Let denote last year's mean cheese consumption by Americans.
(a) H
0
: = 30.0 lb
(b) H
a
: > 30.0 lb
(c) This is a right-tailed test.
16. (a) Rejection region: z 1.28
(e)
(f) Right-tailed test
17. (a) A Type I error would occur if, in fact, = 30.0 lb, but the results
of the sampling lead to the conclusion that > 30.0 lb.
(b) A Type II error would occur if, in fact, > 30.0 lb, but the results
(c) A correct decision would occur if, in fact, = 30.0 lb and the
if, in fact, > 30.0 lb and the results of the sampling lead to that
conclusion.
(d) If, in fact, last year's mean consumption of cheese for all Americans
has not increased over the 2001 mean of 30.0 lb, and we do not reject
the null hypothesis that = 30.0 lb, we made a correct decision.
(e) If, in fact, last year's mean consumption of cheese for all Americans
has increased over the 2001 mean of 30.0 lb, and we fail to reject the
null hypothesis that = 30.0 lb, we made a Type II error.
18. (a) P(Type I error) = significance level = o = 0.10
(b) The distribution of x
_
will be approximately normal with a mean of 30.0
and a standard deviation of 17 . 1 35 / 9 . 6 = .
(c) Note: z
x
n
x z n =
= +
o
o
0
0
/
/
0
if z > 1.28; or
.
So reject H
0
if x
_
0
.
If = 30.5, then
P(Type II error) = P(x
_
< 31.49)
Do not
Reject H
0
Reject H
0
0 1.28 z
0.9000 0.1000
Critical Value
Nonrejection region |Rej ectionreg ion
Power
0.0
0.2
0.4
0.6
0.8
1.0
1.2
30 31 32 33 34 35
Truem ean
P
o
w
e
r
= P(z < (31.49 30.50)/(6.9/ 35 )
= P(z < 0.85) = 0.8023
(d-e) Assuming that the true mean is one of the values listed, the
distribution of x
_
will be approximately normal with that mean and with
a standard deviation of 1.166. The computations of | and the power 1
| are shown in the table below.
computation | 1 |
30.5 85 . 0
35 / 9 . 6
50 . 30 49 . 31
=
0.8023 0.1977
31.0 42 . 0
35 / 9 . 6
00 . 31 49 . 31
=
0.6628 0.3372
31.5 01 . 0
35 / 9 . 6
50 . 31 49 . 31
=
0.4960 0.5040
32.0 44 . 0
35 / 9 . 6
00 . 32 49 . 31
=
0.3300 0.6700
32.5 87 . 0
35 / 9 . 6
50 . 32 49 . 31
=
0.1922 0.8078
33.0 29 . 1
35 / 9 . 6
00 . 33 49 . 31
=
0.0985 0.9015
33.5 72 . 1
35 / 9 . 6
50 . 33 49 . 31
=
0.0427 0.9573
34.0 15 . 2
35 / 9 . 6
00 . 34 49 . 31
=
0.0158 0.9842
(f)
(g) The distribution of x
_
will be approximately normal with a mean of 30.0
and a standard deviation of 6 9 60 0 891 . / . = .
(h) Note: z
x
n
x z n =
= +
o
o
0
0
/
/
0
if z > 1.28; or
equivalently if . 14 . 31 60 / ) 9 . 6 ( 28 . 1 00 . 30 = + > x
So reject H0 if x
_
If = 30.5, then
P(Type II error) = P(x
_
< 31.14)
= P(z < (31.14 - 30.5)/(6.9/ 60 )
= P(z < 0.72) = 0.7642
(i-j)Assuming that the true mean is one of the values listed, the
distribution of x
_
will be approximately normal with that mean and with
a standard deviation of 0.891. The computations of and the power 1
are shown in the following table.
computation | 1 |
30.5 72 . 0
60 / 9 . 6
50 . 30 14 . 31
=
= z 0.7642 0.2358
31.0 16 . 0
60 / 9 . 6
00 . 31 14 . 31
=
= z 0.5636 0.4364
31.5 40 . 0
60 / 9 . 6
50 . 31 14 . 31
=
= z 0.3446 0.6554
32.0 97 . 0
60 / 9 . 6
00 . 32 14 . 31
=
= z 0.1660 0.8340
32.5 53 . 1
60 / 9 . 6
50 . 32 14 . 31
=
= z 0.0630 0.9370
33.0 09 . 2
60 / 9 . 6
00 . 33 14 . 31
=
= z 0.0183 0.9817
33.5 65 . 2
60 / 9 . 6
50 . 33 14 . 31
=
= z 0.0040 0.9960
34.0 21 . 3
60 / 9 . 6
00 . 34 14 . 31
=
= z 0.0007 0.9993
Power
0.0
0.2
0.4
0.6
0.8
1.0
1.2
30 31 32 33 34 35
TrueMe an
P
o
w
e
r
(k)
(l) The principle being illustrated is that increasing the sample size for
a hypothesis test without changing the significance level increases
the power.
19. (a) n = 35, x
_
= 1078/35 = 30.8, o = 6.9
Step 1: H0: = 30.0 lb, Ha: > 30.0 lb
Step 2: o = 0.10
Step 3: z = (30.8 - 30.0)/(6.9/ 35 ) = 0.69
Step 5: Since 0.69 < 1.28, do not reject H0.
sufficient evidence to conclude that last year's mean cheese
consumption for all Americans has increased over the 2001
mean of 30.0 lb.
(b) Given the conclusion in part (a), if an error has been made, it must be
a Type II error. This is because, given that the null hypothesis was
not rejected, the only error that could be made is the error of not
rejecting a false null hypothesis.
20. (a) Step 1: H
0
: = 30.0, H
a
: > 30.0
Step 2: o = 0.10
Step 3: z = 0.69
Step 4: P = 1 - 0.7549 = 0.2451
0
.
sufficient evidence to conclude that last year's mean cheese
consumption for all Americans has increased over the 2001
mean of 30.0 lb.
(b) Using Table 9.12, we classify the strength of evidence against the null
hypothesis as weak or none because P > 0.10.
21. n = 12, x
_
= $284.10, s = $86.90
Step 1: H
0
: = $332, H
a
: < $332
Step 2: o = 0.05
Step 3: t = (284.1 332)/(86.90/ 12 ) = -1.909
Step 5: Since -1.909 < -1.782, reject H
0
.
evidence to conclude that the mean value lost because of purse
snatching has decreased from the 2002 mean of $332.
22. (a) n=12
H0: q = 332; Ha: q < 332
o = 0.05
Critical value = 12(13)/2 - 61 = 17
x
x-
0
q =D
207 -125 125 10 -10
237 -95 95 7 -7
422 90 90 6 6
226 -106 106 9 -9
272 -60 60 3 -3
205 -127 127 11 -11
362 30 30 2 2
348 16 16 1 1
165 -167 167 12 -12
266 -66 66 5 -5
269 -63 63 4 -4
430 98 98 8 8
W = sum of the positive signed ranks = 17
Since 17 is less than or equal to the critical value of 17, we reject
the null hypothesis and conclude that there is evidence that last
years mean value lost to purse snatching has decreased from the 2002
mean.
(b) In performing the Wilcoxon signed-rank test, we are assuming that the
distribution of last years values lost to purse snatching is
symmetric.
(c) If the distribution of values lost is, in fact, a normal distribution,
it is permissible to use the Wilcoxon test since a normal distribution
is also symmetric.
23. If the values lost last year do have a normal distribution, the t-test is
the preferred procedure for performing the hypothesis test since it is
the more powerful test when the distribution is normal, that is, it has
a greater chance of rejecting a false null hypothesis.
24. (a) If the odds-makers are estimating correctly, the mean point-spread
error is zero.
(b) It seems reasonable to assume that the distribution of point spread
errors is approximately normal. In any case, the sample size of 2109
is very large, so the t-test of H0: = 0 vs. Ha: = 0 is
appropriate. At the 5% significance level, the critical values are
1.960. Since ( 0.2 0.0)/(10.9 / 2109) 0.843 t = = , we do not reject H0.
(c) There is not sufficient evidence to conclude that the mean point-spread
is different from zero.
25. Since the distribution is symmetric, use the Wilcoxon signed-rank test. The
sample size is 50 and o is known, so the z-test may be appropriate. Use
caution, however, since it appears that there may be outliers.
26. The distribution is far from normal, being left-skewed. However, since the
sample size is 37 and o is unknown, it is probably reasonable to use the t-
test.
27. (a) The Wilcoxon signed-rank test is appropriate in Exercise 25 since the
distribution is symmetric. It is not appropriate in Exercise 26 since
the distribution is highly skewed to the left.
(b) The sample size is large (n=50) and the distribution is symmetric in
Exercise 26, so either the z-test or the Wilcoxon signed-rank test
could be used. Since the distribution appears to be more peaked with
longer tails than a normal distribution would have, the Wilcoxon test
is preferable.
28. Using Minitab, we choose Stat

Basic Statistics

1-Sample t..., click in
the Samples in columns text box, enter COST in the Samples in columns text
box, click in the Test mean text box and enter 168 in the Test mean text
box, click on the Options button, type 90.0 in the Confidence level text
box, click on the arrow to the right of the Alternative box and select
greater than, and click OK. Then choose Stat

Nonparametrics

1-Sample
Wilcoxon..., enter COST in the Variables text box, click on the Test median
button and type 168 in its text box, select greater than in the Alternative
text box, and click OK. Click on the Graphs button and check the boxes for
Histogram of data and Boxplot of data. Click OK twice. Then choose Stat

Basic Statistics

Normality test, enter COST in the Variable text box and
click OK. Finally, choose Graph

Stem-and-Leaf and enter COST in the
Graph Variables text box and click OK. The results are
(a) Test of mu = 168 vs > 168
95%
Lower
COST 11 185.182 55.496 16.733 154.855 1.03 0.164
The p-value is greater than the significance level of 0.10, so we do not reject
the null hypothesis. There is not sufficient evidence to claim that the
average cost of a private room in a nursing home exceeded $168 per day.
(b) Test of median = 168.0 versus median > 168.0
N
COST 11 11 48.0 0.099 186.8
The p-value is less than the significance level of 0.10, so we reject the null
hypothesis. There is sufficient evidence to claim that the average cost of a
private room in a nursing home exceeded $168 per day.
COST
F
r
e
q
u
e
n
c
y
300 250 200 150 100 50
6
5
4
3
2
1
0
-1
_
X
Ho
Histogramof COST
(with Ho and 95% t-confidence interval for the mean)
COST
300 250 200 150 100
_
X
Ho
Boxplot of COST
COST
P
e
r
c
e
n
t
350 300 250 200 150 100 50
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
0.398
185.2
StDev 55.50
N 11
AD 0.352
P- Value
Probability Plot of COST
Normal
Stem-and-leaf of COST N = 11 Leaf Unit = 10
LO 7
2 1 2
3 1 5
3 1
(5) 1 88899
3 2 0
2 2
2 2 5
HI 28
(d) Although the distribution may be symmetric, it is not bell-shaped and
it contains two potential outliers (see boxplot). Since the t-test
should not be used when the sample size is small (11) and there are
outliers, the Wilcoxon test is the more appropriate one to use.
29. (a,b) Using Minitab, choose Stat

Basic Statistics

1-Sample t, enter
CONSUMPTION in the Samples in columns text box and 64.5 in the Test
mean text box. Click on the Graphs button and check the boxes for
Histogram of data and Boxplot of data. Click OK. Click on the
Options button, enter 95 in the Confidence level text box and select
Less than from the Alternative drop down box. Click OK twice. Then
choose Stat

Basic Statistics

Normality test and enter CONSUMPTION
in the Variable text box and click OK. Finally, choose Graph

Stem-
and-Leaf and enter CONSUMPTION in the Graph Variables text box and
click OK. The results are
CONSUMPTION
F
r
e
q
u
e
n
c
y
80 60 40 20 0
15. 0
12. 5
10. 0
7. 5
5. 0
2. 5
0. 0 _
X
Ho
Histogramof CONSUMPTION
CONSUMPTION
90 80 70 60 50 40 30 20 10 0
_
X
Ho
Boxplot of CONSUMPTION
(with Ho and 95% t-c onfi dence interval for the mean)
CONSUMPTION
P
e
r
c
e
n
t
120 100 80 60 40 20 0
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
<0.005
58.4
StDev 20.42
N 40
AD 2.212
P-V alu e
Probability Plot of CONSUMPTION
Normal
Stem-and-leaf of CONSUMPTION N = 40
Leaf Unit = 1.0
LO 0, 0, 8, 20
5 3 7
5 4
7 4 79
10 5 014
16 5 666667
(7) 6 0112223
17 6 57789
12 7 1234
8 7 5567789
1 8
1 8 9
CONSUMPTION 40 58.4000 20.4172 3.2282 (51.8703, 64.9297) -1.89 0.066
The P-value of the test is 0.066, which is greater than the 0.05
significance level. We do not reject H0. The data do not provide
sufficient evidence to conclude that the mean beef consumption this
year is less than the 2002 mean of 64.5 lbs.
(c) After removing the four outliers, the test results are
CONSUMPTION 36 64.1111 11.0163 1.8360 (60.3837, 67.8385) -0.21 0.833
(d) The outliers had a very large effect on the test results. Although the
sample size was 40, four outliers is too many for the t-test to be
appropriate. If the outliers are not recording errors, they represent
legitimate observations from the population and should not be deleted.
If they were deleted, the test results would not yield valid
conclusions about the population. With so many outliers, it is just
possible that the population itself is quite left skewed as is the
sample.
(e) This is more appropriately done with a non-parametric test that is
insensitive to outliers.
30. (a) Using Minitab, choose Stat

Nonparametrics

1-Wilcoxon, enter
CONSUMPTION in the Samples in columns text box and 64.5 in the Test
mean text box. Click on the Options button, enter 95 in the
Confidence level text box and select Less than from the Alternative
drop down box. Click OK twice. The results are
N
CONSUMPTION 40 40 324.5 0.127 62.00
After removing the outliers and repeating part (a), the results are
N
CONSUMPTION 36 36 324.5 0.450 64.50
(b) The results are similar to those in Problem 29 in that the P-value
increases substantially. The results are different in that the P-
value increases from 0.066 to 0.833 for the t-test, and only from 0.127
to 0.450 for the Wilcoxon test when the four outliers are deleted. In
both cases, we should expect the P-value to increase since the four
smallest observations were deleted and the alternative hypothesis was
< 64.5. At the 5% significance level, the conclusion does not change
in either case.
(c) While the Wilcoxon test is less sensitive to outliers, it is based on
an assumption that the data are symmetrically distributed. That
assumption does not appear to be reasonable for these data, so caution
is advised in using the Wilcoxon test. The sign test would be a better
choice.
31. (a) Using Minitab, choose Stat

Basic Statistics

1-Sample z, enter BMI
in the Samples in columns text box and 25 in the Test mean text box.
Click on the Graphs button and check the boxes for Histogram of data
and Boxplot of data. Click OK. Click on the Options button, enter 95
in the Confidence level text box and select Greater than from the
Alternative drop down box. Click OK twice. Then choose Stat

Basic
Statistics

Normality test and enter BMI in the Variable text box and
click OK. Finally, choose Graph

Stem-and-Leaf and enter BMI in the
Graph Variables text box and click OK. The graphic results are
BMI
F
r
e
q
u
e
n
c
y
35 30 25 20 15
16
12
8
4
0
_
X
Ho
Histogramof BMI
(with Ho and 95% Z-confidence i nterval for the Mean, and StDev = 5)
BMI
40 35 30 25 20 15
_
X
Ho
Boxplot of BMI
BMI
P
e
r
c
e
n
t
45 40 35 30 25 20 15 10
99.9
99
95
90
80
70
60
50
40
30
20
10
5
1
0.1
Mean
0.517
25. 99
StDev 5.029
N 75
A D 0.325
P- Value
Probability Plot of BMI
Normal
Stem-and-leaf of BMI N = 75
Leaf Unit = 1.0
1 1 4
4 1 677
6 1 88
16 2 0000111111
29 2 2222222233333
(9) 2 444444455
37 2 666666777
28 2 88888899999
17 3 0000111
10 3 22233
5 3 4555
1 3 7
(b) Yes. There are no outliers, the sample size is large, and the data is
reasonably normally distributed.
(c) The process in part (a) yielded the following test results:
95%
Lower
BMI 75 25.9867 5.0293 0.5774 25.0370 1.71 0.044
Since the P-value of 0.044 is less than the significance level 0.05, we
reject the null hypothesis. The data do provide sufficient evidence to
conclude that the mean BMI of U.S. adults is greater than that for a
healthy weight.
32. (a) Using Minitab, we choose Graph

Histogram, choose the Simple version,
enter BEER in the Graph variables text box, and click OK. The result
is
BEER
F
r
e
q
u
e
n
c
y
105 90 75 60 45 30 15 0
50
40
30
20
10
0
Histogramof BEER
(b) There may be an outlier at about 105.
(c) Since the sample size is very large (300), we can use the z-test or the
t-test. Since the standard deviation is unknown, we use the t-test.
Using Minitab, we obtained the following result using all of the data.
99%
Lower
BEER 300 27.5200 19.4265 1.1216 24.8967 4.92 0.000
After eliminating the potential outlier (106), we repeated the process
and obtained
99%
Lower
BEER 299 27.2575 18.9187 1.0941 24.6985 4.81 0.000
Clearly, elimination of the value 106 had little effect on the value of t or on
the P-value. In either case, we reject the null hypothesis. The data do
provide sufficient evidence to claim that the mean annual consumption of beer
in Washington, D.C. exceeds the national mean.
Using the Focus Database: Chapter 9
(a) Using Minitab, we choose Stat

Basic Statistics

1-Sample t, enter COMP
in the Samples in columns text box, enter 20.8 in the Test mean text box,
click on the Options button, enter 95 in the confidence level text box, and
select greater than from the Alternative drop down box. Click OK twice. The
results are
Test of mu = 20.8 vs > 20.8
95%
Lower
COMP 200 23.7400 3.1371 0.2218 23.3734 13.25 0.000
Since the P-value is 0.000, we reject the null hypothesis. The data provide
sufficient to claim that the mean ACT score at UWEC is greater than the
national mean of 20.6.
Case Study: Sex and Sense of Direction 457
(b) To obtain the true mean from the FOCUS database, select Stat

Basic
Statistics

Display Descriptive Statistics, enter COMP in the Variables
text box and click OK. The results are
Variable N N* Mean SE Mean StDev Minimum Q1 Median Q3
COMP 6738 0 23.621 0.0385 3.162 13.000 21.000 23.000 26.000
Variable Maximum
COMP 34.000
The mean is shown as 20.621.
(c) Yes. The population mean is, in fact, greater than 20.8. It would not
necessarily have to be correct. Another sample might yield a sample mean
small at random enough to not reject the null hypothesis.
(d-e) Following the same procedure as in part (a) with the variables ENGLISH and
MATH, we get the results shown below
Test of mu = 20.3 vs > 20.3
95%
Lower
ENGLISH 200 23.1700 4.0364 0.2854 22.6983 10.06 0.000
Test of mu = 20.6 vs > 20.6
95%
Lower
MATH 200 23.5650 3.9041 0.2761 23.1088 10.74 0.000
Since the P-values are both 0.000, we reject the null hypotheses. The data
provide sufficient to claim that the mean MATH and ENGLISH scores at UWEC
are greater than the respective national means of 20.3 and 20.6. In both
cases, we see from the entire database that the conclusions are correct.
Case study: Sex and Sense of Direction
(a) If the women just randomly guessed at the direction of south, their mean
pointing error would be zero with a range from -180 to +180 degrees. The
absolute pointing errors would range from 0 to 180 with a mean of 90
degrees.
(b) n = 15, df = 14, x = 55 4 . , s = 42.2
Step 1: H
0
: = 90 H
a
: < 90
Step 2: o = 0.01
Step 4: (55.4 90.0)/ (42.2/ 15) 3.175 t = =
Step 5: Since -3.175 < -2.624, we reject the null hypothesis.
Step 6: We conclude that the women who claim to have a good sense of
direction really do better, on the average, than would be done
by just randomly guessing at the direction of south.
(c) The four plots are
ERROR
F
r
e
q
u
e
n
c
y
120 100 80 60 40 20 0
5
4
3
2
1
0
Histogram of ERROR
E
R
R
O
R
140
120
100
80
60
40
20
0
Boxplot of ERROR
ERROR
P
e
r
c
e
n
t
150 100 50 0 -50
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean
0.071
55.4
St Dev 42.24
N 15
A D 0.652
P- Value
Probability Plot of ERROR
Normal
Stem-and-leaf of ERROR N = 15
Leaf Unit = 10
4 0 0111
(4) 0 2233
7 0
7 0 667
4 0 9
3 1 0
2 1 22
Since this is a small sample and the data do not appear to come from a
normal distribution, it is not reasonable to use a t-test.
(d) With the data in a column named ERROR, choose Stat

Basic Statistics

1-
Sample t..., click in the Samples in columns text box and enter ERROR, click
in the Test mean text box and type 90, click the Options button, enter 99 in
the Confidence level text box, select less than in the Alternative box and
click OK twice. To obtain the boxplot, choose Graph

Boxplot..., select
Simple, click OK, enter ERROR in Graph variables text box, and click OK. To
obtain the histogram, choose Graph

Histogram..., select Simple, click OK,
enter ERROR in the Graph variables text box, and click OK. To obtain the
stem-and-leaf plot, we choose Graph

Stem-and-leaf..., enter ERROR in the
Graph Variables text box, and click OK. The resulting graphs are shown in
part (c). The t-test result is
99%
Upper
ERROR 15 55.4000 42.2371 10.9056 84.0216 -3.17 0.003
(e) n=15
H
0
: = 90; H
a
: < 90
Case Study: Sex and Sense of Direction 459
= 0.01
Critical value = 15(16)/2 - 100 = 20
x
x-0=D D Rank of D Signed Rank R
14 -76 76 13 -13
91 1 1 1 1
27 -67 67 10 -10
122 32 32 6 6
8 -82 82 15 -15
68 -22 22 5 -5
128 38 38 7 7
78 -12 12 2 -2
20 -70 70 11 -11
109 19 19 3 3
31 -59 59 9 -9
69 -21 21 4 -4
12 -78 78 14 -14
36 -54 54 8 -8
18 -72 72 12 -12
W = sum of the positive signed ranks = 17
Since 17 is less than or equal to the critical value of 20, we reject the
null hypothesis and conclude that there is evidence that the women who claim
to have a good sense of direction really do better, on the average, than
would be done by just randomly guessing at the direction of south.
(f) If the women were merely guessing the direction of south at random, their
guesses would be uniformly distributed over the range from -180 to +180
degrees from south and their absolute pointing errors would be uniformly
distributed over the range from 0 to 180 degrees. The normal probability
plot shows a slight s-shaped curve that is also characteristic of a uniform
distribution although the curve might be a little more symmetric if the data
were exactly from a uniform distribution. The other plots show a slight
tendency toward right-skewness, but this could easily be the result of the
small sample size. Overall, the Wilcoxon signed-rank test seems to be a
reasonable choice given both the expectations under the null hypothesis and
the plots.
(g) Choose Stat

Nonparametrics

1-Sample Wilcoxon..., select ERROR in the
Variables text box, click on the Test median button and enter 90 in its text
box, select less than in the Alternative text box, and click OK. The result
is
N for Wilcoxon Estimated
ERROR 15 15 17.0 0.008 52.50
This result is the same as that obtained in part (e).

Chapter 9

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Chapter 9

Transféré par

Droits d'auteur :

Formats disponibles

371

= z 1.000 0.9515 = 0.0485 0.9515

= z 1.000 0.8962 = 0.1038 0.8962

= z 1.000 0.8051 = 0.1949 0.8051

= z 1.000 0.6772 = 0.3228 0.6772

= z 1.000 0.5239 = 0.4761 0.5239

= z 1.000 0.3669 = 0.6331 0.3669

= z 1.000 0.2327 = 0.7673 0.2327

= z 1.000 0.1292 = 0.8708 0.1292

= z 1.000 0.0630 = 0.9370 0.0630

= z 1.000 0.0268 = 0.9732 0.0268

= = 1.000 0.9989 = 0.0011 0.9989

= = 1.000 0.9916 = 0.0084 0.9916

= = 1.000 0.9964 = 0.0036 0.9964

= = 1.000 0.8508 = 0.1492 0.8508

= = 1.000 0.6406 = 0.3594 0.6406

= = 1.000 0.3783 = 0.6217 0.3783

= = 1.000 0.1635 = 0.8365 0.1635

= = 1.000 0.0485 = 0.9515 0.0485

equals 1 since W = 0 or 1 has probability 0.125.

Vous aimerez peut-être aussi