Вы находитесь на странице: 1из 32

Chapter 7 Answer Key

Essentials of Biostatistics for Public Health, 2nd Edition


Lisa M. Sullivan, PhD
Chapter 7 Solutions to Problems
1.

The following data was collected in a clinical trial evaluating a new compound designed
to improve wound healing in trauma patients. The new compound is compared against a
placebo. After treatment for 5 days with the new compound or placebo the extent of wound
healing is measured and the data are shown below.
Treatment
New Compound (n=125)
Placebo (n=125)

0-25%
15
36

Percent Wound Healing


26-50%
51-75%
37
32
45
34

76-100%
41
10

Is there a difference in the extent of wound healing by treatment? (Hint: Are treatment
and the percent wound healing independent?) Run the appropriate test at a 5% level of
significance.
Step 1.

Set up hypotheses and determine level of significance.


H0: Treatment and percent wound healing are independent
H1: H0 is false.
=0.05

Step 2.

Select the appropriate test statistic.


(O - E) 2

E
2

The condition for appropriate use of the above test statistic is that each expected
frequency is at least 5. In Step 4 we will compute the expected frequencies and we
will ensure that the condition is met.
Step 3.

Set up decision rule.


df=(2-1)(4-1)=3 and the decision rule is Reject H0 if 2 > 7.81.

Step 4.

Compute the test statistic.


We now compute the expected frequencies using the formula,
Expected Frequency = (Row Total * Column Total)/N.
The top number in each cell of the table is the observed frequency and the bottom
number is the expected frequency, shown in parentheses.

Treatment
New Compound
Placebo

0-25%
15
(25.5)
36

Percent Wound Healing


26-50%
51-75%
37
32
(41)
(33)
45
34

2012 Jones & Bartlett Learning, LLC

Total
76-100%
41
(25.5)
10

125
125
1

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
(25.5)
51

Total

(41)
82

(33)
66

(25.5)
51

250

The test statistic is computed as follows:


(15 25.5) 2 (37 41) 2 (32 33) 2 (41 25.5) 2
2

25.5
41
33
25.5
2
2
2
(36 25.5)
(45 41)
(34 33)
(10 25.5) 2

25.5
41
33
25.5
2 = 4.32 + 0.39 + 0.03 + 9.42 + 4.32 + 0.39 + 0.03 + 9.42 = 28.32
Step 5.

2.

Conclusion.
Reject H0 because 28.32 > 7.81. We have statistically significant evidence at
=0.05 to show that H0 is false; treatment and percent wound healing are not
independent. The p-value is p < 0.005.

Use the data in Problem 1 and pool the data across the treatments into one sample of size
n=250. Use the pooled data to test whether the distribution of the percent wound healing is
approximately normal. Specifically, use the following distribution: 30%, 40%, 20% and 10%
and =0.05 to run the appropriate test.
Treatment
Number of Patients

0-25%
51

Percent Wound Healing


26-50%
51-75%
82
66

Step 1.

Set up hypotheses and determine level of significance.


H0: p1=0.3, p2=0.4, p3=0.2, p4=0.1
H1: H0 is false.
=0.05

Step 2.

Select the appropriate test statistic.


2

76-100%
51

Total
250

(O - E) 2
E

We must assess whether the sample size is adequate. Specifically, we need to check
min(np0, np1, , n pk) > 5. The sample size here is n=250 and the proportions
specified in the null hypothesis are 0.3, 0.4, 0.2 and 0.1. Thus, min( 250(0.3),
250(0.4), 250(0.2), 250(0.1))=min(75, 100, 50, 25)=25. The sample size is more
than adequate so the formula can be used..
Step 3.

Set up decision rule.


df=k-1=4-1=3, Reject H0 if 2 > 7.81.

2012 Jones & Bartlett Learning, LLC

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 4.

Compute the test statistic.

Treatment
Number of Patients
Expected

0-25%
51
75

Percent Wound Healing


26-50%
51-75%
82
66
100
50

76-100%
51
25

Total
250
250

The test statistic is computed as follows:


(51 75) 2 (82 100) 2 (66 50) 2 (51 25) 2
2

75
100
50
25
2
= 7.68 + 3.24 + 5.12 + 27.04 = 43.08.
Step 5.

3.

Conclusion.
Reject H0 because 43.08 > 7.81. We have statistically significant evidence at
=0.05 to show that H0 is false. The p-value is p < 0.005.

The following table displays descriptive statistics on the participants involved in the
study described in Problem 1.
Characteristic
Mean Age, years
% Males
Mean Educational Level, years
Mean Annual Income, $000s
Mean Body Mass Index

New Compound
47.2
44%
13.1
$36,560
24.7

Placebo
46.1
59%
14.2
$37,470
25.1

p-value
0.7564
0.0215
0.6898
0.3546
0.0851

Are any of the characteristics significantly different between groups? Justify


briefly. (Hint: No calculations, just interpret above.)
There is a statistically significant difference in the proportions of males (p=0.0215) and a
marginally significant difference in mean body mass index (p=0.0851). None of the other
characteristics are significantly different between groups using a 5% level of significance.
4.

An investigator hypothesizes that cholesterol levels in children might be affected by


educating their parents on proper nutrition and exercise. A sample of 40 families with a child
between the ages of 10-15 who has been diagnosed with high cholesterol agree to participate in
the study. All parents are provided educational information on nutrition and exercise. After
following the prescribed program, their childs total cholesterol level is measured. The childrens
mean cholesterol level is 175 with a standard deviation of 19.5. Is there significant evidence of a
reduction in total cholesterol in the children? Run the appropriate test at the 5% level of
significance and assume that the null value for total cholesterol is 191.

2012 Jones & Bartlett Learning, LLC

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 1.

Set up hypotheses and determine level of significance


H0: =
191
H1: <
191
=0.05

Step 2.

Select the appropriate test statistic.


X 0
Z
s/ n .

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Reject H0 if Z < -1.645.

Z
Step 5.

5.

X 0
s/ n

175 - 191
19.5/ 40 = -5.19

Conclusion.
Reject H0 because -5.19 < -1.645. We have statistically significant evidence at
=0.05 to show that the mean total cholesterol level is lower than 191. The p-value
is p<0.0001.

The following data were collected in an experiment designed to investigate the impact of
different positions of the mother during ultrasound on fetal heart rate. Fetal heart rate is
measured by ultrasound in beats per minute. The study includes 20 women who are assigned
to one position and have the fetal heart rate measured in that position. Each woman is
between 28-32 weeks gestation. The data are shown below.
Back
140
144
146
141
139
Mean = 142.0

Side
141
143
145
144
136
Mean = 141.8

Sitting
144
145
147
148
144
Mean = 145.6

Standing
147
145
148
149
145
Mean =146.8

Is there a significant difference in mean fetal heart rates by position? Run the test
at a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: =
2=
3=
4
H1: Means are not all equal
=0.05

Step 2.

Select the appropriate test statistic. F=MSB/MSE.

2012 Jones & Bartlett Learning, LLC

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 3.

Set up decision rule.


df1=k-1=4-1=3 and df2=N-k=20-4=16. Reject H0 if F > 3.24.

Step 4.

Compute the test statistic.


Back
n1=5

Side
n2=5

X 1= 142.0

X 2= 141.8

Sitting
n3=5
X 3= 145.6

Standing
n4=5
X 4= 146.8

If we pool all N=20 observations, the overall mean is X = 144.1.


2

SSB = n j (X j - X )
We can now compute
.
2
2
SSB = 5(142.0-144.1) + 5(141.8-144.1) + 5(145.6 144.1)2 + 5(146.8 -144.1)2
SSB = 96.2
2

( - )
Next, SSE = X X j .
Back
(X 142.0)
140
-2
144
2
146
4
141
-1
139
-3

(X 142.0)2
4
4
16
1
9
34.0

Thus, (X- X 1)2 = 34.0.


Side
141
143
145
144
136

(X 141.8)
-0.8
1.2
3.2
2.2
-5.8

(X 141.8)2
0.64
1.44
10.24
4.84
33.64
50.8

(X 145.6)
-1.6
-0.6
1.4
2.4
-1.6

(X 145.6)2
2.56
0.36
1.96
5.76
2.56
13.2

Thus, (X- X 2)2 = 50.8.


Sitting
144
145
147
148
144

2012 Jones & Bartlett Learning, LLC

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Thus, (X- X 3)2 = 13.2.
Standing
147
145
148
149
145

(X 146.8)2
0.04
3.24
1.44
4.84
3.24
12.8

(X 146.8)
0.2
-1.8
1.2
2.2
-1.8

Thus, (X- X 4)2 = 12.8.


2

SSE = ( X - X j) = 110.8.
We can now construct the ANOVA table.
Source of
Variation

Sums of Squares
SS

Degrees
of
freedom
df

Mean Squares
MS

Between
Treatments

96.2

32.1

4.7

Error or
Residual

110.8

16

6.9

Total

207.0

19

Step 5.

6.

Conclusion.
We reject H0 because 4.7 > 3.24. We have statistically significant evidence at
=0.05 to show that there is a difference in mean fetal heart rate by position.

A clinical trial is conducted comparing a new pain reliever for arthritis to a placebo.
Participants are randomly assigned to receive the new treatment or a placebo and the outcome is
pain relief within 30 minutes. The data are shown below.
New Medication
Placebo

Pain Relief
44
21

No Pain Relief
76
99

Is there a significant difference in the proportions of patients reporting pain relief? Run
the test at a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


2012 Jones & Bartlett Learning, LLC

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
H0: p1 = p2 (equivalent to RD = 0, RR=1 or OR=1)
H1: p1 p2 (equivalent to RD 0, RR 1 or OR 1)
=0.05
Step 2.

Select the appropriate test statistic.


p1 - p 2
Z
1
1
p(1 - p)

n 1 n2

Step 3.

Set up decision rule. Reject H0 if Z < -1.960 or if Z > 1.960.

Step 4.

Compute the test statistic.


We first compute the overall proportion of successes:

x1 x 2
44 21

n 1 n 2 120 120 = 0.27.

We now substitute to compute the test statistic.


p1 - p 2
0.367 0.175
Z

1
1
1
1
0.27(1 0.27)

p(1 - p)

120 120
n1 n 2
Step 5.

=3.35.

Conclusion.
Reject H0 because 3.35 > 1.960. We have statistically significant evidence at
=0.05 to show that there is a difference in the proportions of patients reporting
pain relief. The p-value is p<0.001.

7.

A clinical trial is planned to compare an experimental medication designed to lower


blood pressure to a placebo. Before starting the trial, a pilot study is conducted involving 7
participants. The objective of the study is to assess how systolic blood pressure changes over
time untreated. Systolic blood pressures are measured at baseline and again 4 weeks later. Is
there a statistically significant difference in blood pressures over time? Run the test at a 5% level
of significance.
Baseline
4 Weeks

120
122

145
142

130
135

160
158

152
155

143
140

126
130

Difference
Difference2

-2
4

3
9

-5
25

2
4

-3
9

3
9

-4
16

2012 Jones & Bartlett Learning, LLC

Differences = -6
Differences2 = 76
7

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Xd

sd

76 (6) 2 /7
3.44
6

Step 1.

Set up hypotheses and determine level of significance


H0: d=
0
H1: d
0
=0.05

Step 2.

Select the appropriate test statistic.


X d
t d
sd / n .

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Reject H0 if t < -2.447 or it t > 2.447.

Step 5.

8.

6
0.86
7

Xd d
sd / n

0.86
3.44 / 7 = -0.66.

Conclusion.
Do not reject H0 because -2.447 < -0.66 < 2.447. We do not have statistically
significant evidence at =0.05 to show that there is a difference in blood pressure
over 4 weeks.

The main trial in Problem 7 is conducted and involves a total of 200 patients. Patients are
enrolled and randomized to receive either the experimental medication or the placebo. The data
shown below are data collected at the end of the study after 6 weeks on the assigned treatment.
Mean (SD) Systolic Blood Pressure
% Hypertensive
% With Side Effects

Experimental (n=100)
120.2 (15.4)
14%
6%

Placebo (n=100)
131.4 (18.9)
22%
8%

a. Test if there is a significant difference in mean systolic blood pressures between groups
using =0.05.
Step 1.

Set up hypotheses and determine level of significance


H0: =
2
H1:
2
=0.05

2012 Jones & Bartlett Learning, LLC

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 2.

Select the appropriate test statistic. Check s12/s22 = 15.42/18.92 = 0.66.


X1 - X 2
Z
1
1
Sp

n1 n 2
.

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Sp

Reject H0 if Z < -1.960 or if Z > 1.960.

(100 1)15.4 2 (100 1)18.9 2


100 100 2

=17.23

Now the test statistic,

120.2 - 131.4

17.23
Step 5.

1
1

100 100 = -4.60.

Conclusion.
Reject H0 because -4.60 < -1.960. We have statistically significant evidence at
=0.05 to show that the mean blood pressures are not equal. The p-value is
p<0.0001.

b. Test if there is a significant difference in the proportions of hypertensive patients between


groups using =0.05.
Step 1.

Set up hypotheses and determine level of significance


H0: p1 = p2 (equivalent to RD = 0, RR=1 or OR=1)
H1: p1 p2 (equivalent to RD 0, RR 1 or OR 1)
=0.05

Step 2.

Select the appropriate test statistic.


p1 - p 2
Z
1
1
p(1 - p)

n 1 n2

Step 3.

Set up decision rule. Reject H0 if Z < -1.960 or if Z > 1.960.

Step 4.

Compute the test statistic.


We first compute the overall proportion of successes:
2012 Jones & Bartlett Learning, LLC

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD

x1 x 2
14 22

n 1 n 2 100 100 = 0.18.

We now substitute to compute the test statistic.


p1 - p 2
0.14 0.22
Z

1
1
1
1
0.18(1 0.18)

p(1 - p)

100 100
n1 n 2
Step 5.

9.

= -1.47.
Conclusion.
Do not reject H0 because -1.960 < -1.47 < 1.960. We do not have statistically
significant evidence at =0.05 to show that there is a difference in the proportions
of patients with hypertension.

Suppose in the trial described in Problem 8 that patients were recruited from 3 different
clinical sites. Use the following data to test if there is a difference in the proportions of
hypertensive patients across clinical sites.
Hypertensive
Not Hypertensive

Site 1
10
68

Site 2
14
56

Step 1.

Set up hypotheses and determine level of significance.


H0: Site and hypertensive status are independent
H1: H0 is false.
=0.05

Step 2.

Select the appropriate test statistic.


2

Site 3
12
40

(O - E) 2
E

The condition for appropriate use of the above test statistic is that each expected
frequency is at least 5. In Step 4 we will compute the expected frequencies and we
will ensure that the condition is met.
Step 3.

Set up decision rule.


df=(2-1)(3-1)=2 and the decision rule is Reject H0 if 2 > 5.99.

Step 4.

Compute the test statistic.


We now compute the expected frequencies using the formula,
Expected Frequency = (Row Total * Column Total)/N.

2012 Jones & Bartlett Learning, LLC

10

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
The top number in each cell of the table is the observed frequency and the bottom
number is the expected frequency, shown in parentheses.
Hypertensive
Not Hypertensive
Total

Site 1
10
(14.0)
68
(64.0)
78

Site 2
14
(12.6)
56
(57.4)
70

Site 3
12
(9.4)
40
(42.6)
52

Total
36
164
200

The test statistic is computed as follows:


(10 14.0) 2 (14 12.6) 2 (12 9.4) 2
2

14.0
12.6
9.4
2
2
(68 64.0)
(56 57.4)
(40 42.6) 2

64.0
57.4
42.6
2 = 1.14 + 0.15 + 0.72 + 0.25 + 0.03 + 0.16 = 2.46.
Step 5.

10.

Conclusion.
Do not reject H0 because 2.46 < 5.99. We do not have statistically significant
evidence at =0.05 to show that H0 is false or that site and hypertensive status are
not independent.

A clinical trial is conducted to compare an experimental medication to placebo to reduce


the symptoms of asthma. Two hundred participants are enrolled in the study and randomized to
receive either the experimental medication or placebo. The primary outcome is self-reported
reduction of symptoms. Among 100 participants who receive the experimental medication, 38
report a reduction of symptoms as compared to 21 participants of 100 assigned to placebo. Test
if there is a significant difference in the proportions of participants reporting a reduction of
symptoms between the experimental and placebo groups. Use =0.05.
Step 1.

Set up hypotheses and determine level of significance


H0: p1 = p2 (equivalent to RD = 0, RR=1 or OR=1)
H1: p1 p2 (equivalent to RD 0, RR 1 or OR 1)
=0.05

Step 2.

Select the appropriate test statistic.


p1 - p 2
Z
1
1
p(1 - p)

n 1 n2

2012 Jones & Bartlett Learning, LLC

11

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 3.

Set up decision rule. Reject H0 if Z < -1.960 or if Z > 1.960.

Step 4.

Compute the test statistic.


We first compute the overall proportion of successes:

x1 x 2
38 21

n 1 n 2 100 100 = 0.295.

We now substitute to compute the test statistic.


p1 - p 2
0.38 0.21
Z

1
1
1
1
0.295(1 0.295)

p(1 - p)

100
100

n
n
2
1
Step 5.

11.

= 2.64.

Conclusion.
Reject H0 because 2.64 > 1.960. We have statistically significant evidence at
=0.05 to show that there is a difference in the proportions of patients reporting a
reduction in symptoms. The p-value is p<0.010.

Suppose more detail is actually recorded in the primary outcome in the clinical trial
described in Problem 10. The data are recorded as follows.
Treatment
Experimental
Placebo

Much
Worse
10
12

Worse

Change in Symptoms
No change
Better

17
25

35
42

28
12

Much
Better
10
9

Is there a difference in change in symptoms by treatment group? Run the appropriate test
at a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance.


H0: Treatment and change in symptoms are independent
H1: H0 is false.
=0.05

Step 2.

Select the appropriate test statistic.


2

2012 Jones & Bartlett Learning, LLC

(O - E) 2
E

12

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
The condition for appropriate use of the above test statistic is that each expected
frequency is at least 5. In Step 4 we will compute the expected frequencies and we
will ensure that the condition is met.
Step 3.

Set up decision rule.


df=(2-1)(5-1)=4 and the decision rule is Reject H0 if 2 > 9.49.

Step 4.

Compute the test statistic.


We now compute the expected frequencies using the formula,
Expected Frequency = (Row Total * Column Total)/N.
The top number in each cell of the table is the observed frequency and the bottom
number is the expected frequency, shown in parentheses.
Treatment
Experimental
Placebo
Total

Much
Worse
10
(11)
12
(11)
22

Change in Symptoms
Worse
No
Better
change
17
35
28
(21)
(38.5)
(20)
25
42
12
(21)
(38.5)
(20)
42
77
40

Much
Better
10
(9.5)
9
(9.5)
19

Total
100
100
200

The test statistic is computed as follows:


(10 11) 2 (17 21) 2 (35 38.5) 2 (28 20) 2 (10 9.5) 2
2

11
21
38.5
20
9.5
2
2
2
2
(12 11)
(25 21)
(42 38.5)
(12 20)
(9 9.5) 2

11
21
38.5
20
9.5
2 = 0.09 + 0.76 + 0.32 + 3.20 + 0.03 + 0.09 + 0.76 + 0.32 + 0.03 = 8.8.
Step 5.

12.

Conclusion.
Do not reject H0 because 8.8 < 9.49. We do not have statistically significant
evidence at =0.05 to show that H0 is false or that treatment and change in
symptoms are not independent.

Suppose a secondary outcome is recorded in the trial described in Problem 10 reflecting


asthma symptom severity measured on a scale of 0-100 with higher scores indicating more
severe symptoms. In the participants who receive the experimental medication the mean
symptom score is 74 with a standard deviation of 5.6 and in the placebo group the mean
symptom score is 85 with a standard deviation of 6.0. Is there a significant difference in mean
symptom scores between groups? Run the appropriate test at a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


2012 Jones & Bartlett Learning, LLC

13

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
H0: =
2
H1:
2

=0.05

Step 2.

Select the appropriate test statistic. Check s12/s22 = 5.62/6.02 = 0.87.


X1 - X 2
Z
1
1
Sp

n1 n 2
.

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Sp

Reject H0 if Z < -1.960 or if Z > 1.960.

(100 1)5.6 2 (100 1)6.0 2


100 100 2

=5.8

Now the test statistic,

74 - 85

Z
5.8
Step 5.

13.

1
1

100 100 = -13.41.

Conclusion.
Reject H0 because -13.41 < -1.960. We have statistically significant evidence at
=0.05 to show that the mean symptom scores are not equal. The p-value is
p<0.0001.

Recent recommendations suggest 60 minutes of physical activity per day. A sample of 50


adults in a study of cardiovascular risk factors report exercising a mean of 38 minutes per day
with a standard deviation of 19 minutes. Based on the sample data, is the physical activity
significantly less than recommended? Run the appropriate test at a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: =
60
H1: <
60
=0.05

Step 2.

Select the appropriate test statistic.


X 0
Z
s/ n .

2012 Jones & Bartlett Learning, LLC

14

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 3.

Set up decision rule.

Reject H0 if Z < -1.645.

Step 4.

Compute the test statistic.

Z
Step 5.

14.

X 0
s/ n

38 - 60
19/ 50 = -8.19.

Conclusion.
Reject H0 because -8.19 < -1.645. We have statistically significant evidence at
=0.05 to show that the mean number of minutes of exercise is less than 60. The pvalue is p<0.0001.

Suppose a hypertension trial is mounted and 18 participants are randomly assigned to one
of the comparison treatments. Each participant takes the assigned medication and their systolic
blood pressure (SBP) is recorded after 6 months on the assigned treatment. The data are as
follows.
Standard Treatment
124
111
133
125
128
115

Placebo
134
143
148
142
150
160

New Treatment
114
117
121
124
122
128

Is there a difference in mean SBP among treatments? Run the appropriate test at
=0.05.
Step 1.

Set up hypotheses and determine level of significance


H0: =
2=
3
H1: Means are not all equal
=0.05

Step 2.

Select the appropriate test statistic. F=MSB/MSE.

Step 3.

Set up decision rule.


df1=k-1=3-1=2 and df2=N-k=18-3=15. Reject H0 if F > 3.68.

Step 4.

Compute the test statistic.

Standard
n1=6
X 1= 122.7

Placebo
n2=6
X 2= 146.2

2012 Jones & Bartlett Learning, LLC

New Treatment
n3=6
X 3= 121.0
15

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
If we pool all N=18 observations, the overall mean is X = 130.0.
SSB = n j (X j - X )

We can now compute


.
SSB = 6(122.7-130.0)2 + 6(146.2-130.0)2 + 6(121.0-130.0)2
SSB = 2380.4.
2
( - )
Next, SSE = X X j .
Standard Treatment
124
111
133
125
128
115

(X 122.7)
1.3
-11.7
10.3
2.3
5.3
-7.7

(X 122.7)2
1.69
136.89
106.09
5.29
28.09
59.29
337.34

(X 146.2)
-12.2
-3.2
1.8
-4.2
3.8
13.8

(X 146.2)2
148.84
10.24
3.24
17.64
14.44
190.44
384.84

(X 121.0)
-7
-4
0
3
1
7

(X 121.0)2
49
16
0
9
1
49
124

Thus, (X- X 1)2 = 337.34.


Placebo
134
143
148
142
150
160
Thus, (X- X 2)2 = 384.84.
New Treatment
114
117
121
124
122
128
Thus, (X- X 3)2 = 124.0
2

SSE = ( X - X j) = 846.18.
We can now construct the ANOVA table.
2012 Jones & Bartlett Learning, LLC

16

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD

Source of
Variation

Sums of Squares
SS

Degrees
of
freedom
df

Mean Squares
MS

Between
Treatments

2380.4

1190.2

21.1

Error or
Residual

846.2

15

56.4

Total

3226.6

17

Step 5.

15.

Conclusion.
We reject H0 because 21.1 > 3.68. We have statistically significant evidence at
=0.05 to show that there is a difference in mean systolic blood pressure among
treatments.

A study is conducted to compare mean cholesterol levels for individuals following a low
carbohydrate diet for at least 6 months to individuals following a conventional (low-fat, low
calorie) diet for at least 6 months. The data are summarized below:
Diet Program
Low Carbohydrate
Conventional

Sample Size
50
75

Mean Cholesterol
225.4
203.8

Std Dev Cholesterol


24.5
21.6

Test if there is a significant difference in mean cholesterol levels between the diet
programs using a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: =
2
H1:
2
=0.05

Step 2.

Select the appropriate test statistic. Check s12/s22 = 24.52/21.62 = 1.29.


X1 - X 2
Z
1
1
Sp

n1 n 2
.

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Reject H0 if Z < -1.960 or if Z > 1.960.

2012 Jones & Bartlett Learning, LLC

17

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD

Sp

(50 1)24.5 2 (75 1)21.6 2


50 75 2

= 22.8.

Now the test statistic,


Z

225.4 - 203.8
22.8

Step 5.

16.

1
1

50 75 = 5.19.

Conclusion.
Reject H0 because 5.19 > 1.960. We have statistically significant evidence at
=0.05 to show that the mean cholesterol levels are not equal. The p-value is
p<0.0001.

Another outcome variable in the study described in Problem 15 was hypercholesterolemia, defined as total cholesterol over 220. Among the individuals who follow
the low carbohydrate diet, 56% are hypercholesterolemic and among the individuals who
follow the conventional diet, 40% are hypercholesterolemic. Test if there is a significant
difference in the proportions using a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: p1 = p2 (equivalent to RD = 0, RR=1 or OR=1)
H1: p1 p2 (equivalent to RD 0, RR 1 or OR 1)
=0.05

Step 2.

Select the appropriate test statistic.


p1 - p 2
Z
1
1
p(1 - p)

n 1 n2

Step 3.

Set up decision rule. Reject H0 if Z < -1.960 or if Z > 1.960.

Step 4.

Compute the test statistic.


We first compute the overall proportion of successes:
x1=0.56(50) = 28 and x2=0.40(75) = 30.

x 1 x 2 28 30

n 1 n 2 50 75 = 0.46.

2012 Jones & Bartlett Learning, LLC

18

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
We now substitute to compute the test statistic.
p1 - p 2
0.56 0.40
Z

1
1
1
1
0.46(1 0.46)

p(1 - p)

50 75
n1 n 2
Step 5.

= 1.76.

Conclusion.
Do not reject H0 because -1.960 < 1.76 < 1.960. We do not have statistically
significant evidence at =0.05 to show that there is a difference in the proportions
of patients with hypercholesterolemia between treatments.
The following table compares background characteristics of the participants
involved in the study described in Problem 15. Are there any statistically
significant differences in patient characteristics between the different diet
programs? Justify briefly.
Low carbohydrate
(n=50)
52.1
42
15.3
$39,540

Mean age (years)


Men (%)
Mean educational level (years)
Mean family income

Conventional
(n=75)
53.4
34
12.9
$47,980

p-value
0.7564
0.0145
0.0237
0.0576

There is a statistically significant difference in the proportions of men in each diet group
(p=0.0145) and in the mean educational levels between diet groups (p=0.0237). There is
also a marginally statistically significant difference in mean family income (p=0.0576.
17.
18.

MISSING
Suppose the results of the analysis in Problems 15 to Problem 17 are reported and
criticized because the participants were not randomized to different diets and that there may be
other factors associated with changes in cholesterol. A third study is run to estimate the effect of
the low carbohydrate diet on cholesterol levels. In the third study, participants cholesterol levels
are measured before starting the program and then again after 6 months on the program. The
data are shown below.
Before Program
After 6 Months

210
215

230
240

190
190

215
200

260
280

200
210

Is there a significant increase in cholesterol after 6 months on the low carbohydrate diet?
Run the appropriate test at a 5% level of significance.

2012 Jones & Bartlett Learning, LLC

19

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Difference
Difference2

5
25

10
100

0
0

-15
225

Xd

30
5.0
6

20
400

10
100

Differences = 30
Differences2 = 850

850 (30) 2 /6
sd
11.8
5
Step 1.

Set up hypotheses and determine level of significance


H0: d=
0
H1: d>
0
=0.05

Step 2.

Select the appropriate test statistic.


X d
t d
sd / n .

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Reject H0 if t > 2.015.

Step 5.

19.

Xd d
sd / n

5 .0
11.8 / 6 = 1.04.

Conclusion.
Do not reject H0 because 1.04 < 2.015. We do not have statistically significant
evidence at =0.05 to show that there is a difference in cholesterol after 6 months
on the diet.

A study is conducted to compare 3 new appetite suppressants (A, B, C) to a placebo in


terms of their effects on weight reduction. A total of 80 participants are involved and are
randomly assigned to the comparison groups (in equal numbers). The outcome of interest is
weight reduction, measured in pounds. The following data are observed after 3 months on
treatment.
Mean (SD)
Weight Reduction

A
6.4 (4.1)

B
8.9 (4.0)

C
2.2 (3.9)

Placebo
2.5 (4.3)

Is there a significant difference in mean weight reduction among the 4 treatments? Use a
5% level of significance. Hint: SST=1,889.

2012 Jones & Bartlett Learning, LLC

20

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 1.

Set up hypotheses and determine level of significance


H0: =
2=
3=
4
H1: Means are not all equal
=0.05

Step 2.

Select the appropriate test statistic. F=MSB/MSE.

Step 3.

Set up decision rule.


df1=k-1=4-1=3 and df2=N-k=80-4=76. Reject H0 if F > 2.73.

Step 4.

Compute the test statistic.


To organize our computations we will complete the ANOVA table. In order to
compute the sums of squares we must first compute the sample means for each
group and the overall mean.
A
B
C
Placebo
n1=20
n2=20
n3=20
n4=20
X 1= 6.4
X 2= 8.9
X 3= 2.2
X 4= 2.5
If we pool all N=80 observations, the overall mean is X = 5.0.
SSB = n j (X j - X )

We can now compute


.
SSB = 20(6.4 5.0)2 + 20(8.9-5.0)2 + 20(2.2-5.0)2 + 20(2.5-5.0)2
SSB = 39.2 + 304.2 + 156.8 + 125.0 = 625.2

Source of
Variation

Sums of Squares
SS

Degrees
of
freedom
df

Mean Squares
MS

Between
Treatments

625.2

208.4

12.53

Error or
Residual

1263.8

76

16.63

Total

1889.0

79

Step 5.

20.

Conclusion.
Reject H0 because 12.53 > 2.73. We have statistically significant evidence at
=0.05 to show that there is a difference in mean weight reduction among the four
treatments.

The mean lifetime for cardiac stents is 8.9 years. A medical device company has
implemented some improvements in the manufacturing process and hypothesizes that the
2012 Jones & Bartlett Learning, LLC

21

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
lifetime is now longer. A study of 40 new devices reveals a mean lifetime of 9.7 years with a
standard deviation of 3.4 years. Is there statistical evidence of a prolonged lifetime of the stents?
Run the test at a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: =
8.9
H1: >
8.9
=0.05

Step 2.

Select the appropriate test statistic.


X 0
Z
s/ n .

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Reject H0 if Z > 1.645.

Z
Step 5.

21.

X 0
s/ n

9.7 - 8.9
3.4/ 40 = 1.49

Conclusion.
Do not reject H0 because 1.49 < 1.645. We do not have statistically significant
evidence at =0.05 to show that the lifetime is longer than 8.9 years.
A study is conducted in 100 children to assess risk factors for obesity. Children are
enrolled and undergo a complete physical examination. At the examination, height and
weight of the child, their mother and father are measured and converted to body mass
index scores (weight(kg)/height(m)2). Data on self-reported health behaviors are
captured by interview and merged with the physical examination data. The primary
outcome variable is childs obesity for analysis, children are classified as normal weight
(BMI<25) or overweight/obese (BMI>25). Data on key study variables are summarized
below.

Characteristics
Mean (SD) Age, years
% Male
Mothers BMI
Normal (BMI<25)
Overweight (BMI 25-29.9)
Obese (BMI>30)
Fathers BMI
Normal (BMI<25)

Normal Weight

Overweight/Obese

Total

(n=62)
13.4 (2.6)
45%

(n=38)
11.1 (2.9)
51%

(n=100)
12.5 (2.7)
47%

40 (65%)
15 (24%)
7 (11%)

16 (41%)
14 (38%)
8 (21%)

56 (56%)
29 (29%)
15 (15%)

34 (55%)

16 (41%)

50 (50%)

2012 Jones & Bartlett Learning, LLC

22

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Overweight (BMI 25-29.9)
Obese (BMI>30)
Mean (SD) Systolic Blood

20 (32%)
8 (13%)
123 (15)

14 (38%)
8 (21%)
139 (12)

34 (34%)
16 (16%)
129 (14)

Pressure
Mean (SD) Total Cholesterol

186 (25)

211 (28)

196 (26)

Is there a statistically significant difference in mean age between normal and


overweight/obese children? Run the test at a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: =
2
H1:
2
=0.05

Step 2.

Select the appropriate test statistic. Check s12/s22 = 2.62/2.92 = 0.80.


X1 - X 2
Z
1
1
Sp

n1 n 2
.

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Sp

Reject H0 if Z < -1.960 or if Z > 1.960.

(62 1)2.6 2 (38 1)2.9 2


62 38 2

= 2.7.

Now the test statistic,

13.4 - 11.1
2.7

Step 5.

22.

1
1

62 38 = 4.13.

Conclusion.
Reject H0 because 4.13 > 1.960. We have statistically significant evidence at
=0.05 to show that the mean ages are not equal. The p-value is p<0.0001.
Use the data shown in Problem 21 and test if there is an association between mothers
BMI and the childs obesity status (i.e., normal versus overweight/obese)? Run the test at
a 5% level of significance.

2012 Jones & Bartlett Learning, LLC

23

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 1.

Set up hypotheses and determine level of significance.


H0: Mothers BMI and childs obesity status are independent
H1: H0 is false.
=0.05

Step 2.

Select the appropriate test statistic.


2

(O - E) 2
E

The condition for appropriate use of the above test statistic is that each expected
frequency is at least 5. In Step 4 we will compute the expected frequencies and we
will ensure that the condition is met.
Step 3.

Set up decision rule.


df=(3-1)(2-1)=2 and the decision rule is Reject H0 if 2 > 5.99.

Step 4.

Compute the test statistic.


We now compute the expected frequencies using the formula,
Expected Frequency = (Row Total * Column Total)/N.
The top number in each cell of the table is the observed frequency and the bottom
number is the expected frequency, shown in parentheses.

Mother:
Normal
Overweight
Obese
Total

Child:
Normal
40
(34.7)
15
(18.0)
7
(9.3)
62

Child:
Overwt/Obese
16
(21.3)
14
(11.0)
8
(5.7)
38

Total
56
29
15
100

The test statistic is computed as follows:


(40 34.7) 2 (16 21.3) 2 (15 18.0) 2 (14 11.0) 2 (7 9.3) 2 (8 5.7) 2
2

34.7
21.3
18.0
11.0
9 .3
5 .7
2
= 0.81 +1.32 + 0.5 + 0.82 + 0.57 + 0.93 = 4.95.
Step 5.

Conclusion.
Do not reject H0 because 4.95 < 5.99. We do not have statistically significant
evidence at =0.05 to show that H0 is false or that mothers BMI and childs
obesity status are not independent.

2012 Jones & Bartlett Learning, LLC

24

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
23.

Use the data shown in Problem 21 and test if there is a significant difference in the
proportions normal versus overweight/obese children who are male. Run the test at a 5%
level of significance.

Step 1.

Set up hypotheses and determine level of significance


H0: p1 = p2 (equivalent to RD = 0, RR=1 or OR=1)
H1: p1 p2 (equivalent to RD 0, RR 1 or OR 1)
=0.05

Step 2.

Select the appropriate test statistic.


p1 - p 2
Z
1
1
p(1 - p)

n 1 n2

Step 3.

Set up decision rule. Reject H0 if Z < -1.960 or if Z > 1.960.

Step 4.

Compute the test statistic.

p
The overall proportion is given as:

x1 x 2
n 1 n 2 = 0.47.

We now substitute to compute the test statistic.


p1 - p 2
0.45 0.51
Z

1
1
1
1
0.47(1 0.47)

p(1 - p)

62 38
n1 n 2
Step 5.

24.

= -0.58.

Conclusion.
Do not reject H0 because -1.96 < -0.58 < 1.96. We do not have statistically
significant evidence at =0.05 to show that there is a difference in the proportions
of normal versus overweight/obese children who are male.

A study is run to compare body mass index (BMI) in participants assigned to different
diet programs and the data are analyzed in Excel. Use the Excel results to answer the questions
below.
a.

Complete the following ANOVA Table

ANOVA
Source of Variation

SS

df

2012 Jones & Bartlett Learning, LLC

MS

P-value

F crit
25

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Between Groups
Within Groups
Total

40.791667
35.166666
75.958327

3
20
23

13.597
1.758

7.734

0.0012733

3.10

b. Write the hypotheses to be tested.


H0: =
2=
3=
4
H1: Means are not all equal

=0.05

c. Write the conclusion of the test.


Reject H0 because 7.734 > 3.10 (p=0.0013). We have significant evidence,
=0.05, that the means are not all equal
25.

A clinical trial is conducted to test the efficacy of a new drug for hypertension. The new
drug is compared to a standard drug and to a placebo in a study involving n=120
participants. The primary outcome is systolic blood pressure measured after 4 weeks on
the assigned drug. The table below shows characteristics of study participants measured
at baseline (prior to randomization).

Baseline Characteristic
Mean (SD) Age
N (%) Male
Mean (SD) BMI
Mean (SD) SBP
N (%) Diabetic
N (%) Current Smokers

Placebo (n=40)
75.2 (4.4)
12 (30.0%)
26.1 (4.9)
142.1 (19.2)
13 ( 32.5%)
4 (10.0%)

Standard Drug (n=40)


75.6 (4.8)
13 (32.5%)
25.1 (3.0)
150.4 (19.8)
11 (27.5%)
2 (5.0%)

New Drug (n=40)


74.7 (5.6)
9 (22.5%)
26.9 (4.1)
144.5 (19.7)
8 (20.0%)
1 (2.5%)

Test if the mean SBP at baseline is significantly higher in participants assigned to the
Standard Drug as compared to those assigned to the New Drug. Use a 5% level of
significance.
Step 1.

Set up hypotheses and determine level of significance


H0: S=
N
H1: S
N
=0.05

Step 2.

Select the appropriate test statistic. Check s12/s22 = 19.82/19.72 = 1.01.


X1 - X 2
Z
1
1
Sp

n1 n 2
.

2012 Jones & Bartlett Learning, LLC

26

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Sp

Reject H0 if Z > 1.645.

(40 1)19.8 2 (40 1)19.7 2


40 40 2

= 19.75.

Now the test statistic,


150.4 - 144.5

19.75
Step 5.

26.

1
1

40 40 = 1.34

Conclusion.
Do not reject H0 because 1.34 < 1.645. We do not have statistically significant
evidence at =0.05 to show that the mean SBP at baseline is significantly higher
in participants assigned to the Standard Drug as compared to those assigned to the
New Drug.
Use the data shown in Problem 25 and test if there is a significant difference in the
proportions of men assigned to each of the 3 treatments. (HINT: Are gender and
treatment independent?) Use a 5% level of significance.

Step 1.

Set up hypotheses and determine level of significance.


H0: Gender and treatment group are independent
H1: H0 is false.
=0.05

Step 2.

Select the appropriate test statistic.


2

(O - E) 2
E

The condition for appropriate use of the above test statistic is that each expected
frequency is at least 5. In Step 4 we will compute the expected frequencies and we
will ensure that the condition is met.
Step 3.

Set up decision rule.


df=(2-1)(3-1)=2 and the decision rule is Reject H0 if 2 > 5.99.

Step 4.

Compute the test statistic.


We now compute the expected frequencies using the formula,
Expected Frequency = (Row Total * Column Total)/N.
2012 Jones & Bartlett Learning, LLC

27

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
The top number in each cell of the table is the observed frequency and the bottom
number is the expected frequency, shown in parentheses.

Male
Female
Total

Placebo
12
(11.3)
28
(28.7)
40

Standard
13
(11.3)
27
(28.7)
40

New
9
(11.3)
31
(28.7)
40

Total
34
86
120

The test statistic is computed as follows:


(12 11.3) 2 (13 11.3) 2 (9 11.3) 2 (28 28.7) 2 (27 28.7) 2 (31 28.7) 2
2

11.3
11.3
11.3
28.7
28.7
28.7
2 = 0.04 +0.26 + 0.47 + 0.02 + 0.10 + 0.18 = 1.07.
Step 5.

27.

Conclusion.
Do not reject H0 because 1.07 < 5.99. We do not have statistically significant
evidence at =0.05 to show that H0 is false or that gender and treatment group are
not independent.

Use the data shown in Problem 25 and test if a there is a significant difference in the
proportion of diabetic participants in the Placebo as compared to the Standard Drug groups. Use
a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: pP = pS (equivalent to RD = 0, RR=1 or OR=1)
H1: pP pS (equivalent to RD 0, RR 1 or OR 1)
=0.05

Step 2.

Select the appropriate test statistic.


p1 - p 2
Z
1
1
p(1 - p)

n 1 n2

Step 3.

Set up decision rule. Reject H0 if Z < -1.960 or if Z > 1.960.

Step 4.

Compute the test statistic.

p
The overall proportion is given as:
2012 Jones & Bartlett Learning, LLC

x 1 x 2 13 11

n1 n 2
80 = 0.3.
28

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
We now substitute to compute the test statistic.
p1 - p 2
0.325 0.275
Z

1
1
1
1
0.3(1 0.3)

p(1 - p)

40 40
n1 n 2
Step 5.

28.

= 0.49.

Conclusion.
Do not reject H0 because -1.96 < 0.49 < 1.96. We do not have statistically
significant evidence at =0.05 to show that there is a difference in the proportions
of diabetic participants in the Placebo as compared to the Standard Drug groups.

Use the data shown in Problem 25 and test if there is a significant difference in mean age
among the 3 groups. (HINT: SStotal=2893.) Use a 5% level of significance.
Step 1.

Set up hypotheses and determine level of significance


H0: =
2=
3
H1: Means are not all equal
=0.05

Step 2.

Select the appropriate test statistic. F=MSB/MSE.

Step 3.

Set up decision rule.


df1=k-1=3-1=2 and df2=N-k=120-3=117. Reject H0 if F > 3.09.

Step 4.

Compute the test statistic.


To organize our computations we will complete the ANOVA table
The overall mean is X = (75.2+75.6+74.7)/3 = 75.2.
2

SSB = n j (X j - X )
We can now compute
.
2
2
SSB = 40(75.2 75.2) + 40(75.6-75.2) + 40(74.7-75.2)2
SSB = 0 + 6.4 + 10.0 = 16.4

Source of
Variation

Sums of Squares
SS

Degrees
of
freedom
df

Mean Squares
MS

Between
Treatments

16.4

8.2

0.33

Error or
Residual

2876.6

117

24.6

2012 Jones & Bartlett Learning, LLC

29

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Total
Step 5.

29.

2893

119

Conclusion.
Do not reject H0 because 0.33 < 3.09. We do not have statistically significant
evidence at =0.05 to show that there is a difference in mean age among the 3
groups.

Some scientists believe that alcoholism is linked to social isolation. One measure of
social isolation is marital status. A study of 280 adults is conducted and each participant
is classified as not alcoholic, diagnosed alcoholic or undiagnosed alcoholic, and
categorized by marital status. Is there significant evidence of an association? Run the
appropriate test at a 5% level of significance.

Married
Not Married

Diagnosed
Alcoholic
21
59

Undiagnosed
Alcoholic
37
63

Step 1.

Set up hypotheses and determine level of significance.


H0: Marital status and alcoholism are independent
H1: H0 is false.
=0.05

Step 2.

Select the appropriate test statistic.


2

Not Alcoholic
58
42

(O - E) 2
E

The condition for appropriate use of the above test statistic is that each expected
frequency is at least 5. In Step 4 we will compute the expected frequencies and we
will ensure that the condition is met.
Step 3.

Set up decision rule.


df=(2-1)(3-1)=2 and the decision rule is Reject H0 if 2 > 5.99.

Step 4.

Compute the test statistic.


We now compute the expected frequencies using the formula,
Expected Frequency = (Row Total * Column Total)/N.
The top number in each cell of the table is the observed frequency and the bottom
number is the expected frequency, shown in parentheses.
Diagnosed
Alcoholic
2012 Jones & Bartlett Learning, LLC

Undiagnosed
Alcoholic

Not Alcoholic

Total
30

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD
Married
Not Married
Total

21
(33.1)
59
(46.9)
80

37
(41.4)
63
(58.6)
100

58
(41.4)
42
(58.6)
100

116
164
280

The test statistic is computed as follows:


(21 33.1) 2 (37 41.4) 2 (58 41.4) 2 (59 46.9) 2 (63 58.6) 2 (42 58.6) 2
2

33.1
41.4
41.4
46.9
58.6
58.6
2 = 4.42 +0.47 + 6.66 +3.12 +0.33 +4.70 =19.7.
Step 5.

30.

Conclusion.
Reject H0 because 19.7>5.99. We have statistically significant evidence at =0.05
to show that H0 is false or that marital status and alcoholism are not independent.
The p-value is p<0.005.

A study is performed to examine the relationship between the concentration of plasma


antioxidant vitamins and cancer risk. The table below shows data for plasma vitamin-A
concentration in stomach-cancer patients and in controls (participants similar to the cancer
patients but free of disease). Is there a significant difference in the mean concentration of plasma
antioxidant vitamins between patients with stomach cancer and controls? Run the appropriate
test at a 5% level of significance.
N
20
50

Stomach cancer patients


Controls

Mean
2.41
2.78

Std Dev
0.15
0.19

Step 1.

Set up hypotheses and determine level of significance


H0: =
2
H1:
2
=0.05

Step 2.

Select the appropriate test statistic. Check s12/s22 = 0.152/0.192 = 0.62.


X1 - X 2
t
1
1
Sp

n1 n 2
, df = n1 + n2 -2 = 20 = 50 2 = 68

Step 3.

Set up decision rule.

Step 4.

Compute the test statistic.

Reject H0 if t < -1.995 or if t > 1.995.

2012 Jones & Bartlett Learning, LLC

31

Chapter 7 Answer Key


Essentials of Biostatistics for Public Health, 2nd Edition
Lisa M. Sullivan, PhD

Sp

(20 1)0.15 2 (50 1)0.19 2


20 50 2

= 0.18.

Now the test statistic,


2.41 - 2.78

0.18
Step 5.

1
1

20 50 = -7.77.

Conclusion.
Reject H0 because -7.77 < -1.960. We have statistically significant evidence at
=0.05 to show that there a significant difference in the mean concentration of
plasma antioxidant vitamins between patients with stomach cancer and controls.
The p-value is p<0.01.

2012 Jones & Bartlett Learning, LLC

32