MINITAB 14 Supplement For Biostatistics For Health Sciences
MINITAB 14 Supplement For Biostatistics For Health Sciences
MINITAB 14 Supplement For Biostatistics For Health Sciences
0
. Now type .8 in the Power values: window and 4 in the Standard
deviation: window. Click the Options... button and be sure the Not equal
button is selected. Type .01 in the Signicance level: window and click OK
twice. MINITAB prints the sample size of 47 in the Session Window which is
the solution calculated in t he text.
4.4 Condence Intervals (page 137)
Task 4.13
Use MINITAB to construct the 95 percent condence interval outlined in Ex-
ample 4.37 on page 143.
Solution
Select StatBasic Statistics1-Sample Z.... Select the Summarized
data button in the 1-Sample Z (Test and Condence Interval) window.
Type 60 in the Sample size: window, 90.1 in the Mean: window and 16
in the Standard deviation: window. Now click the Options... button. Be
4.4. CONFIDENCE INTERVALS 43
sure that the Condence level: is set to 95.0 and the Alternative: is set to
not equal. Click OK twice. The interval, L = 86.0515 and U = 94.1485 are
provided in the Session Window. Both of these values agree with the results
provided in the text.
Task 4.14
Use MINITAB to construct the condence interval described in Example 4.41
on page 147.
Solution
Enter the data from Example 4.41 into C1 of a new worksheet. Select StatBasic
Statistics1-Sample t.... Select the Samples in columns: button and type
C1 in the accompanying window. Now click the Options... button and type
99.0 in the Condence level: window. Be sure not equal is selected in the
Alternative: window and click OK twice. MINITAB gives the condence
interval as (94.654, 120.146) which agrees with the result given in the text.
Task 4.15
Use MINITAB to form the exact and approximate condence intervals con-
structed in Example 4.42 on page 150.
Solution
Select StatBasic Statistics1 Proportion.... Select the Summarized
data button. Enter 10 in the Number of trials: window and 4 in the number
of event: window. Click the Options... button. Be sure that 95.0 is entered in
the Condence level: window and not equal is selected for the Alternative:
window and click OK twice. MINITAB displays the condence interval as
(0.121552, 0.737622) which, given rounding, agrees with the exact computations
carried out in the text.
To obtain the approximate result, repeat the above steps but check the Use
test and interval based on normal distribution box under the Options...
button. MINITAB gives the interval (0.096364, 0.703636) which, again, agrees
with the approximate interval given in the text. It is noteworthy that MINITAB
provides the following admonition in association with this result. * NOTE *
The normal approximation may be inaccurate for small samples.
44 CHAPTER 4. INFERENCE AND ONE SAMPLE METHODS
Exercises
4.1 Use MINITAB to reproduce the binomial probabilities in Table 4.5 on text
page 109 on page 143.
4.2 In Example 4.8 on page 86, the solution to the problem involves nding the
area of the normal curve that lies above a Z value of 1.36. The solution
gives this area as .0895. Use MINITAB and, if necessary a calculator, to
nd this area.
4.3 Use MINITAB to perform the one mean Z test outlined in Example 4.11
on page 95 as well as to nd critical Z for the test.
4.4 Use MINITAB to perform the one mean Z test outlined in Example 4.13
on page 98 as well as to nd critical Z values for the test.
4.5 Use MINITAB to perform the one mean t test outlined in Example 4.16 on
page 106 as well as to nd critical t for the test.
4.6 Use MINITAB to perform the test outlined in Example 4.18 on page 111.
4.7 Use MINITAB to perform the test outlined in Example 4.20 on page 114.
4.8 Use MINITAB to perform the following test of hypothesis using both the
method outlined in the text (i.e. Test One and Test Two) and the shortcut
(i.e. condence interval) method.
H
0E
: .45 or .55
H
AE
: .45 < < .55
where p = .47 and n = 500. Use = .05.
4.9 In reference to Example 4.32, calculate the portion of the alternative curve
that lies in the right hand critical region of the null curve thereby reconcil-
ing the dierence of approximately .0069975 between the power calculation
provided by MINITAB and that calculated in the text.
4.10 Use MINITAB to obtain a good estimate of the exact power calculated in
Example 4.35.
4.11 Use MINITAB to form the condence interval described in Example 4.38
on page 144.
4.12 Use MINITAB to form the condence intervals described in Example 4.39
on page 144.
4.13 Use MINITAB to construct the condence interval outlined in Exercise 4.22
on page 157.
4.14 Use MINITAB to construct the two-sided and one-sided condence intervals
alluded to in Example 4.43 on page 151.
Chapter 5
Paired Samples Methods
5.1 Introduction
In this chapter you will use MINITAB to perform tests of hypotheses and form
condence intervals for various paired samples methods. You will conduct paired
samples t tests, McNemars test, as well as tests on risk and odds ratios. You will
also form condence intervals for each of these statistics and perform equivalence
tests.
5.2 Methods Related to Mean Dierence (page
160)
Task 5.1
Use MINITAB to perform the paired samples t test described in Example 5.1
on page 162.
Solution
Load the worksheet containing the Table 5.1 data by selecting FileOpen
Worksheet... and opening le table 5.1 data.MTW. Now select StatBasic
StatisticsPaired t.... In the Paired t (Test and Condence Interval)
window be sure the Samples in columns button is selected. In the First
sample: window type C2 then type C1 in the Second sample: window. The
sample designations are done in this fashion because MINITAB subtracts the
second sample from the rst sample when performing computations. Now click
the Options... button and be sure that Test mean: is set to 0.0 and Alter-
native: is set to not equal. Click OK twice. MINITAB reports obtained t as
T-Value = 2.93 which is the value computed in the text. The p-value is given
as 0.011 which is less than = .05 so that the null hypothesis is rejected.
45
46 CHAPTER 5. PAIRED SAMPLES METHODS
Task 5.2
Use MINITAB to perform the equivalence test described in Example 5.4 on page
168.
Solution
Load the worksheet containing the Table 5.5 data by selecting FileOpen
Worksheet... and opening le table 5.5 data.MTW. The one-tailed equiva-
lence test may be carried out by conducting a one-tailed paired samples t test as
follows. Select StatBasic StatisticsPaired t.... In the Paired t (Test
and Condence Interval) window be sure the Samples in columns button
is selected. Type C2 in the First sample: window and C1 in the Second
sample: window. Now click the Options... button. Type 6.0 in the Test
mean: window and select less than in the Alternative: window and click OK
twice. MINITAB gives obtained t as 2.61 which agrees with the value calcu-
lated in the text. The p-value is given as 0.009 which results in rejection of the
null hypothesis.
Task 5.3
Use the paired data in Table 5.1 to construct a one-sided 95 percent condence
interval for the lower bound of
d
.
Solution
Load the worksheet containing the Table 5.1 data by selecting FileOpen
Worksheet... and opening le table 5.1 data.MTW. Now select StatBasic
StatisticsPaired t.... In the Paired t (Test and Condence Interval)
window be sure the Samples in columns button is selected. In the First
sample: window type C2 then type C1 in the Second sample: window. Now
click the Options... button and be sure that Condence level: is set to 95.0
and Alternative: is set to greater than.
1
Click OK twice. MINITAB reports
the lower bound estimate as 2.55343 which is the value reported in the solution
to Example 5.6.
5.3 Methods Related to Proportions (page 174)
Task 5.4
Use both approximate and exact methods to perform a two-tailed McNemars
test on the data in Table 5.7 on page 175 at the = .05 level.
1
Read the material under One-Tailed Hypothesis Tests And One-SidedCondence Intervals
on page 154 if you are not sure why the greater than option was chosen.
5.3. METHODS RELATED TO PROPORTIONS 47
Figure 5.1: Cross tabulation of
disease status for vaccine treated
twins.
Solution
Unfortunately, MINITAB does not provide a direct method for performing this
test from the paired data. However, recalling that McNemars test is simply a
test for a proportion, we can use this facility of MINITAB to perform the test.
We begin by nding the number of noninformative pairs, the number of pairs
in which the vaccine one treated twin had disease but the vaccine two treated
twin did not and the number of pairs in which the vaccine one treated twin had
no disease while the vaccine two treated twin did develop the disease.
To begin, load the worksheet containing the Table 5.7 data by selecting
FileOpen Worksheet... and opening le table 5.7 data.MTW. Columns
C1 and C2 contain the paired designations as to whether the twin contracted the
disease (D) or did not contract the disease (D bar) for the two vaccine groups.
Now select StatTablesCross Tabulation and Chi-Square.... Double-
click on Vaccine Two to place this variable in the For rows: window (or
alternatively, type C2 in the window) then double-click Vaccine One to place
it in the For columns: window (or alternatively, type C1 in the window). Put
a check in the Counts window and be sure the other items under Display are
unchecked. Click OK. The output is shown in Figure 5.1.
Notice that for two pairs of twins both had the disease while for six pairs
neither had the disease. Thus, there are eight noninformative pairs. Of the
remaining eight pairs, there were six for whom the Vaccine One twin had disease
while the Vaccine Two twin was disease free. We now have sucient information
to conduct the test.
We are now ready to perform McNemars test. To this end, select StatBasic
Statistics1 Proportion.... Choose the Summarized data: button then
type 8 in the Number of trials: window and 6 in the Number of events:
window. Now click the Options... button and be sure that .5 is specied in
the Test proportion: window and not equal is specied in the Alternative:
window. In order to perform the approximate test we place a check in the Use
test and interval based on normal distribution window and click OK
twice. MINITAB gives obtained Z as 1.41 which is the value computed in the
48 CHAPTER 5. PAIRED SAMPLES METHODS
text on page 176.
The exact test is performed by following the above steps with the exception
that the Use test and interval based on normal distribution window is
not checked. In this case MINITAB gives the two-tailed p-value as .289 which is
the value calculated in conjunction with Example 5.11 on page 179 of the text.
Task 5.5
Use the Outcome variable in Table 5.14 to perform the exact test outlined in
Example 5.13 on page 183.
Solution
Enter the Outcome variable, minus the noninformative observations, from Table
5.14 into C1 of a new worksheet. Select StatBasic Statistics1 Proportion....
Select the Samples in columns: button and type C1 in the accompanying
window. Click the Options... button and enter .4 in the Test proportion:
window. Select greater than in the Alternative: window and click OK twice.
MINITAB gives the p-value for the test as 0.514 which is the value calculated
in the solution to Example 5.13.
Task 5.6
Construct the condence intervals described in Example 5.15 on page 186.
Solution
Select StatBasic Statistics1 Proportion.... Select the Summarized
data button and enter 10 in the Number of trials: window and 4 in the
Number of events: window. Click on Options... and be sure that the Con-
dence level: window contains 95.0. Enter not equal in the Alternative:
window and check the Use test and interval based on normal distribu-
tion: box to obtain the approximate interval. Now click OK twice. MINITAB
displays the interval as (0.096364, 0.703636) which agrees with the calculations
obtained in the solution to Example 5.15.
Repeat the above steps with the Use test and interval based on nor-
mal distribution: box unchecked. MINITAB gives the exact interval as
(0.121552, 0.737622 which, except for rounding, again agrees with the calcu-
lations provided in the solution to 5.15.
5.4 Methods Related to Paired Samples Risk
Ratios (page 190)
Task 5.7
Carry out the tasks outlined in Example 5.17 on page 191.
5.5. METHODS RELATED TO PAIRED SAMPLES ODDS RATIOS 49
Solution
MINITAB does not provide a direct means for calculating a paired samples risk
ratio. However, it can still be useful in such an endeavor in that it can be used
to form the data into a two by two table thereby facilitating the calculation.
This can be particularly useful for large data sets.
For the problem at hand, generate the table shown in Figure 5.1 of this man-
ual by following the steps given in Task 5.4. You can then use the information
given in this table to calculate the risk ratio via Equation 5.7 on page 191.
The test of signicance conducted via Equation 5.4 was carried out in the
solution to Task 5.4. To obtain the result provided by Equation 5.5, simply
square the result given by Equation 5.4.
Task 5.8
Use MINITAB to perform the equivalence test as outlined in Example 5.20 on
page 195.
Solution
Unfortunately, MINITAB does not provide a straightforward means for per-
forming this test.
Task 5.9
Use MINITAB to form the condence interval described in Example 5.22 on
page 197.
Solution
Unfortunately, MINITAB does not provide a straightforward means for con-
structing this interval.
5.5 Methods Related to Paired Samples Odds
Ratios (page 199)
Task 5.10
Use MINITAB to test the hypothesis H
0
: OR = 1 as described in Example
5.24 on page 201.
Solution
Using the fact that a test of H
0
: OR = 1 is equivalent to a test of H
0
: = .5, we
conduct the test as follows. Select StatBasic Statistics1 Proportion....
Choose the Summarized data button and enter 30 in the Number of trials:
window and 25 in the Number of events: window. The number of trials is
specied as 30 because after eliminating the noninformative pairs there are
50 CHAPTER 5. PAIRED SAMPLES METHODS
25 + 5 = 30 remaining pairs that do provide information. Of these, in twenty-
ve pairs, the member with cancer had been exposed while the pair member
without cancer was not exposed. We dene this as the number of events.
Now click Options... and be sure that the Test proportion: is set to 0.5.
Put a check in the Use test and interval based on normal distributionbox
and click OK twice. MINITAB performs the test as a Z test while a chi-square
statistic is reported in the text. By squaring obtained Z of 3.65 we obtain,
except for rounding, the chi-square statistic of 13.33 reported in the text.
Task 5.11
Perform the exact two-tailed equivalence test described in Example 5.27 on page
206.
Solution
Unfortunately, MINITAB does not provide a straightforward means of perform-
ing this test. However, by using the relationship between the paired samples
odds ratio and proportions along with the so called shortcut method of perform-
ing equivalence tests introduced on page 39, the task can be performed without
too much diculty.
We begin by constructing a 90 (not 95) percent condence interval for the
estimation of . To this end, select StatBasic Statistics1 Proportion....
Choose the Summarized data button and enter 15 in the Number of trials:
window and 8 in the Number of events: window. Now click Options...
and be sure that the Condence level: is set to .90 and Alternative: is
not equal. Because we are conducting an exact test, be sure the Use test
and interval based on normal distribution box is not checked. Click OK
twice. The condence interval (0.299986, 0.756273) is displayed in the Session
Window. Because this condence interval is not contained in the equivalence
interval EI
L
= .454 and EI
U
= .545, the equivalence null hypothesis is not
rejected. This is the same conclusion reached in the Solution to Example 5.27.
Task 5.12
Construct the exact condence interval described in Example 5.30 on page 211.
Solution
Unfortunately, MINITAB does not provide a straightforward means of con-
structing this interval. However, by using the relationship between the paired
samples odds ratio and proportions, we can construct the CI for the associated
proportion and then convert this interval for the estimation of into an interval
for estimating the paired samples OR.
To this end, select StatBasic Statistics1 Proportion.... Choose the
Summarized data button and enter 24 in the Number of trials: window and
13 in the Number of events: window. Now click Options... and be sure that
the Condence level: is set to .95 and Alternative: is not equal. Because
5.5. METHODS RELATED TO PAIRED SAMPLES ODDS RATIOS 51
we are conducting an exact test, be sure the Use test and interval based
on normal distribution box is not checked. Click OK twice. The condence
interval (0.328208, 0.744470) is displayed in the Session Window. This is the
same interval reported as part of the solution to Example 5.30. In order to
convert this interval into an interval for the estimation of OR, you must apply
Equation 5.17 (page 208) as is shown in the solution to Example 5.30.
52 CHAPTER 5. PAIRED SAMPLES METHODS
Exercises
5.1 Use MINITAB to perform the test of hypothesis outlined in Example 5.2
on page 163.
5.2 Use MINITAB to perform the equivalence test outlined in Example 5.5 on
page 169.
5.3 Use the paired data in Table 5.2 to form a 95 percent condence interval
for the estimation of
1
2
where
1
represents the mean for Treatment
One and
2
represents the mean for Treatment Two.
5.4 Use MINITAB to perform the test outlined in Example 5.12 on page 180.
5.5 Use the Outcome variable in Table 5.16 to perform the tests alluded to in
Example 5.14 on page 185.
5.6 Use MINITAB to construct the condence intervals discussed in Example
5.16 on page 188.
5.7 Conduct the test outlined in Example 5.18.
5.8 Conduct the test of signicance alluded to in Example 5.19 on page 192.
5.9 Use MINITAB to perform the exact two-tailed test of signicance outlined
in Example 5.25 on page 202.
5.10 Use MINITAB to construct the approximate condence interval discussed
in Example 5.28 on page 208.
Chapter 6
Two Independent Samples
Methods
6.1 Introduction
In this chapter you will use MINITAB to perform tests of hypotheses and form
condence intervals for various two (independent) sample methods. You will
conduct independent samples t tests and tests for dierences between propor-
tions. You will also form condence intervals for each of these statistics and
perform equivalence tests.
6.2 Methods Related to Dierences Between Means
(page 215)
Task 6.1
Conduct the independent samples t test described in Example 6.1 on page 220.
Solution
Open the table 6.1 data.MTW worksheet. Select StatBasic Statistics2-
Sample t.... The 2-Sample t (Test and Condence Interval) window
opens. Select the Samples in dierent columns button. Enter C1 in the
First: window and C2 in the Second: window. Put a check in the Assume
equal variances box. Click the Options... button. Be sure that the Test
dierence: window is set to 0.0 and the Alternative: window is set to not
equal. Click OK twice. The result shown in Figure 6.1 appears in the Session
Window.
Notice that obtained t is 0.81 which is the value calculated in the text. The
associated p-value is 0.426 which results in failure to reject the null hypothesis
53
54 CHAPTER 6. TWO INDEPENDENT SAMPLES METHODS
Figure 6.1: MINITAB output for an independent samples t
test.
at the .05 level.
Task 6.2
Conduct the equivalence test described in Example 6.3 on page 225.
Solution
Open the table 6.1 data.MTW worksheet. Select StatBasic Statistics2-
Sample t.... The 2-Sample t (Test and Condence Interval) window
opens. Select the Samples in dierent columns button. Enter C1 in the
First: window and C2 in the Second: window. Put a check in the Assume
equal variances box. Click the Options... button. Set the Test dierence:
window to 5.0 and the Alternative: window to greater than. Click OK twice.
The result appears in the Session Window. Notice that obtained t of 0.38 is the
same value calculated in the text. The associated p-value is 0.353 so that the
equivalence null hypothesis is not rejected.
Task 6.3
Form the condence interval described in Example 6.5 on page 228.
Solution
Open the table 6.1 data.MTW worksheet. Select StatBasic Statistics2-
Sample t.... The 2-Sample t (Test and Condence Interval) window
opens. Select the Samples in dierent columns button. Enter C1 in the
First: window and C2 in the Second: window. Put a check in the Assume
equal variances box. Click the Options... button. Be sure the Condence
level: is set to 95.0 and the Alternative: is set to not equal. Click OK twice.
MINITAB prints the following in the Session Window.
95% CI for dierence: (-12.02002, 5.22002)
This is the interval calculated in the text.
6.3. METHODS RELATED TO PROPORTIONS 55
6.3 Methods Related to Proportions (page 230)
Task 6.4
Perform the two sample test for a dierence between proportions addressed in
Example 6.7 on page 233.
Solution
Select StatBasic Statistics2 Proportions.... In the 2 Proportions
(Test and Condence Interval) window, choose the Summarized data
button. In the First: window enter 314 for Trials: and 23 for Events:. In
the Second: window enter 316 for Trials: and 39 for Events:. Now click the
Options... button. Be sure that Test dierence: is set to 0.0, Alternative:
is not equal and the Use pooled estimate of p for test box is not checked.
Now click OK twice. MINITAB reports obtained Z as 2.12, which is the result
reported in the text, with an associated p-value of 0.034. This value would be
appropriate for a two-tailed test. For the one-tailed test addressed here we
calculate .034/2 = .017 which causes the null hypothesis to be rejected.
Task 6.5
Perform the two-tailed equivalence test described in Example 6.9 on page 235.
Solution
For Test One, select StatBasic Statistics2 Proportions.... In the 2
Proportions (Test and Condence Interval) window, choose the Sum-
marized data button. In the First: window enter 100 for Trials: and 7 for
Events:. In the Second: window enter 100 for Trials: and 6 for Events:.
Now click the Options... button. Set Test dierence: to 0.04, Alternative:
to less than and be sure the Use pooled estimate of p for test box is not
checked. Now click OK twice. MINITAB reports obtained Z as .86, which is
the result reported in the text, with an associated p-value of 0.195.
For Test Two, follow the same steps as for Test One with the exception that
Test dierence: is set to -.04 and Alternative: to greater than. MINITAB
gives obtained Z as 1.43, which is the value calculated in the text, with an
associated p-value of 0.076. Thus, the equivalence null hypothesis is not rejected.
Task 6.6
Construct the condence interval described in Example 6.11 on page 236.
Solution
Select StatBasic Statistics2 Proportions.... In the 2 Proportions
(Test and Condence Interval) window, choose the Summarized data
button. In the First: window enter 299 for Trials: and 106 for Events:.
56 CHAPTER 6. TWO INDEPENDENT SAMPLES METHODS
In the Second: window enter 313 for Trials: and 66 for Events:. Now
click the Options... button. Be sure that Condence level: is set to 95.0,
Alternative: is not equal and the Use pooled estimate of p for test box
is not checked. Now click OK twice. MINITAB reports the condence interval
as (0.0730676, 0.214237). Notice that these values dier slightly from those
calculated in the text. This is because the methods used by MINITAB and the
text for constructing such intervals dier slightly. The method used in the text
is slightly more conservative and, in the opinion of the authors, is preferable.
6.4 Methods Related to Independent Samples
Risk Ratios (page 238)
Unfortunately, MINITAB does not implement the methods discussed in the text
for dealing with independent samples risk ratios.
6.5 Methods Related to Independent Samples
Odds Ratios (page 247)
Unfortunately, MINITAB does not implement the methods discussed in the text
for dealing with independent samples odds ratios.
Exercises 57
Exercises
6.1 The MINITAB output associated with Task 6.1 contains the statement
Both use Pooled StDev = 11.5245. How does the value 11.5245 relate
to the calculations carried out in the text in conjunction with Example 6.1?
6.2 Use MINITAB to conduct the test of hypothesis described in Example 6.2.
6.3 Perform the equivalence test described in Example 6.4 on page 226.
6.4 Find the condence interval discussed in Example 6.6 on page 229.
6.5 Perform the test described in Example 6.8 on page 233.
6.6 Perform the equivalence test outlined in Example 6.10 on page 235.
6.7 Construct the one-sided condence interval described in Example 6.12 on
page 237.
58 CHAPTER 6. TWO INDEPENDENT SAMPLES METHODS
Chapter 7
Multi-Sample Methods
7.1 Introduction
In this chapter you will use MINITAB to perform a One-Way ANOVA, a 2 by
k chi-square test and Tukeys HSD test.
7.2 The One-Way Analysis of Variance (ANOVA)
F Test (page 264)
Task 7.1
Use the data in Table 7.1 to compute MS
w
, MS
b
, obtained F and to produce
an ANOVA table.
Solution
Load the table 7.1b data.MTW worksheet. Notice that C2 contains the
weights of the 15 subjects while C1 contains an indicator (1, 2, or 3) as to
which group the weight represents. Select StatANOVAOne-Way.... The
One-Way Analysis of Variance window opens. Type C2 in the Response:
window and C1 in the Factor: window. Click OK. The relevant portion of the
MINITAB output is shown in Figure 7.1.
In the ANOVA table provided by MINITAB, between groups elements are
labeled Factor while within group elements are labeled Error. From the
table we see that MS
w
and MS
b
are given as 625 and 935 respectively. Except
for rounding, these are the values calculated in the text. Obtained F is given
as 1.50 which is the value given in the text. The associated p-value is .263.
59
60 CHAPTER 7. MULTI-SAMPLE METHODS
Figure 7.1: MINITAB rendering of ANOVA
table for data in text Table 7.1.
Figure 7.2: Nutritional data ar-
ranged for chi-square analysis.
7.3 The 2 By k Chi-Square Test (page 276)
Task 7.2
Perform the chi-square test outlined in Example 7.6 on page 280.
Solution
Open a new worksheet. Label C1 Instruction and C2 No Instruction. Now
enter the number of Low Birth Weight babies for the Instruction group in R1C1
and the number of No Low Birth Weight babies in R2C1. Enter the correspond-
ing values for the No Instruction group in R1C2 and R2C2. Your worksheet
should appear as in Figure 7.2.
To perform the analysis, select StatTablesChi-Square Test (Table
in Worksheet).... Enter C1 C2 in the Columns containing the table:
window and click OK. The output is shown in Figure 7.3. Notice that in addi-
tion to obtained chi-square with its associated p-value, the output also shows
the observed and expected frequencies for each cell along with the contribution
made by that cell to the overall chi-square value.
1
1
That is, the values summed in the calculation of obtained chi-square.
7.4. MULTIPLE COMPARISON PROCEDURES 61
Figure 7.3: Chi-square analysis of nutritional data.
7.4 Multiple Comparison Procedures (page 282)
Task 7.3
Perform the Tukeys HSD tests outlined in Example 7.9 on page 289.
Solution
Load the table 7.1b data.MTW worksheet. Notice that C2 contains the
weights of the 15 subjects while C1 contains an indicator (1, 2, or 3) as to
which group the weight represents. Select StatANOVAOne-Way.... The
One-Way Analysis of Variance window opens. Type C2 in the Response:
window and C1 in the Factor: window. Click the Comparisons... button.
Put a check in the Tukeys, family error rate: box and enter .05 in the
accompanying window. Click OK twice. In addition to the ANOVA table, the
output shown in Figure 7.4 appears in the Session Window.
Notice that comparisons are made in terms of condence intervals rather
than hypothesis tests as is done in the text. Thus, the comparison
2
1
2
is represented by the interval L = 34.35, U = 49.95. Because zero is in the
interval, the null hypothesis is not rejected. This is true of the remaining com-
parisons as well so that the conclusions provided by MINITAB are the same as
those reached in the text.
2
This comparison is represented as
1
2
in the text.
62 CHAPTER 7. MULTI-SAMPLE METHODS
Figure 7.4: Results of Tukeys HSD tests conducted on the dieting study data
in Table 7.1.
Exercises 63
Exercises
7.1 Carry out the ANOVA analysis outlined in Example 7.3 on page 269.
7.2 Carry out the ANOVA analysis outlined in Example 7.4 on page 271.
7.3 Perform the chi-square analysis outlined in Example 7.5 on page 279.
7.4 Perform the multiple comparison tests outlined in Example 7.10 on page
290.
64 CHAPTER 7. MULTI-SAMPLE METHODS
Chapter 8
The Assessment of
Relationships
8.1 Introduction
In this chapter you will use MINITAB to calculate the Pearson product-moment
correlation coecient and to test the hypothesis H
0
: = 0. Additionally, you
will conduct the chi-square test for independence.
8.2 The Pearson Product-Moment Correlation
Coecient (page 295)
Task 8.1
Calculate the Pearson product-moment correlation coecient for the data in
Table 8.1 on page 297. Perform a two-tailed test of the hypothesis H
0
: = 0
at the .05 level.
Solution
Load the table 8.1 data.MTW worksheet. Select StatBasic StatisticsCorrelation....
Enter x and y in the Variables: window and be sure the Display p-values
box is checked. Click OK. MINITAB displays the correlation as 0.964 which is
the value calculated in the text. The associated p-value of 0.000 results in reject
of the null hypothesis H
0
: = 0.
65
66 CHAPTER 8. THE ASSESSMENT OF RELATIONSHIPS
Figure 8.1: Chi-square analysis of data de-
scribed in Example 8.7.
8.3 The Chi-Square Test For Independence (page
312)
Task 8.2
Perform the chi-square test described in Example 8.7 on page 314.
Solution
Enter the observed frequencies into a new worksheet as shown in manual Fig-
ure 8.1. Then select StatTablesChi-Square Test (Table in Work-
sheet).... Enter C1 C2 C3 in the Columns containing the table: window
and click OK. The observed and expected frequencies for each cell of the table
are printed along with the contribution made by each cell to the overall chi-
square statistic. Obtained chi-square of 206.811 is reported with 4 degrees of
freedom and an associated p-value 0.000.
Exercises 67
Exercises
8.1 Calculate the correlation coecient discussed in Example 8.2 on page 298.
68 CHAPTER 8. THE ASSESSMENT OF RELATIONSHIPS
Chapter 9
Linear Regression
9.1 Introduction
In this chapter, you will use MINITAB to construct simple and multiple regres-
sion models, generate
R
2
values and conduct tests of signicance.
9.2 Simple Linear Regression (page 320)
Task 9.1
Construct the simple linear regression model alluded to in Example 9.1 on page
321.
Solution
Open the table 8.2 data worksheet. Select StatRegressionRegression....
In the Regression window, type C2 in the Response: window and C1 in the
predictors: window. Click Options... and be sure there is a check in the Fit
intercept box. Click OK twice. MINITAB gives the regression equation as
y = 0.874 + 0.706 x which is the result given in the text.
Task 9.2
Find SS
reg
, SS
res
, and
R
2
for the data in Table 9.1 on page 323.
Solution
Open the table 9.1 data worksheet. Select StatRegressionRegression....
In the Regression window, type C1 in the Response: window and C2 in the
predictors: window. Click Options... and be sure there is a check in the
Fit intercept box. Click OK twice. MINITAB gives the Regression sum of
squares (SS
reg
) as 107.44 and the Residual Error sum of squares (SS
res
) as
69
70 CHAPTER 9. LINEAR REGRESSION
8.29 which agrees with the values calculated in the text.
R
2
is given as R-Sq =
92.8% which is the value calculated in the solution to Example 9.2 on page 324.
Task 9.3
Test the null hypothesis R
2
= 0 for a model used to predict y from x for the
data in Table 9.1 on page 323.
Solution
Using the output from the solution to Task 9.2, we see that obtained F is
reported as 168.49 with an associated p-value of 0.000. The value of 168.49
reported by MINITAB diers slightly from the value of 167.56 calculated in the
text. This small dierence appears to be the result of dierences in rounding.
9.3 Multiple Linear Regression (page 329)
Task 9.4
Use the data in Table 9.2 on page 330 to construct a two predictor model with
x
1
and x
2
being used to predict y.
Solution
Open the table 9.2 data worksheet. Select StatRegressionRegression....
In the Regression window, type C1 in the Response: window and C2 C3 in
the predictors: window. Click Options... and be sure there is a check in
the Fit intercept box. Click OK twice. MINITAB gives the regression model
as y = 2.69 + 0.404 x1 + 1.00 x2 which agrees with the results calculated in the
text.
Task 9.5
Test the null hypothesis H
0
:
1
=
2
= 0 for the model you constructed in
Task 9.4.
Solution
From the output for the solution to Task 9.4, we note that obtained F for
the regression model is 4.73 which is the value calculated in the text. The
associated p-value is 0.031 which results in rejection of the null hypothesis for
a test conducted at = .05. This is the same conclusion reached in the text.
Task 9.6
Use the data in Table 9.2 on page 330 to determine whether adding x
2
to a
model that contains x
1
adds signicantly to
R
2
.
9.3. MULTIPLE LINEAR REGRESSION 71
Figure 9.1: Sequential General Linear Model analysis of
Table 9.2 data.
Solution
This problem can be approached in a number of dierent ways. For example,
we could use MINITAB to nd
R
2
y.1
and
R
2
y.12
then use a calculator to apply
Equation 9.24. We will use a more direct, if somewhat less obvious, approach.
Select StatANOVAGeneral Linear Model.... The General Linear
Model screen opens. Type y in the Responses: window and x1 x2 in the
Model: window. To this point we have set up a model in which y is the
dependent variable with x1 and x2 being predictors. Now click on Covariates...
and enter x1 x2 in the Covariates: window. Among other things, this last
action cause MINITAB to treat x1 and x2 as continuous variables rather than
transforming them as might be done in other forms of analysis. Click OK. Now
click Options.... In the Sum of Squares: column, darken the Sequential
(Type I) button. Click OK twice. The relevant portion of the output is shown
in manual Figure 9.1. This output will require some explanation.
Of interest for the problem at hand are the headings labeled Seq SS (Se-
quential Sum of Squares), F (obtained F), and P (p-value). The Seq SS
values represent the increment in SS
reg
that occurs as each variable is added to
the model.
1
Thus, with only x
1
in the model, SS
reg
was 150.83. Adding x
2
to
the model increased this value by 90.78. It will be instructive to convert these
values into expressions of R
2
. This can be done by dividing each value by SS
y
which is labeled in the printout as Total. For x
1
we have
R
2
=
150.83
548.40
= .2750
which is equal to
R
2
y.1
. Of more interest is the
R
2
value obtained from x
2
which
is
R
2
=
90.78
548.40
= .1655
which is equal to
R
2
y.12
R
2
y.1
. The values of F reported in the output are the
values of obtained F for a test of signicance on each of the incremental sums of
squares. It follows that obtained F of 3.55 is obtained F for a test conducted on
1
The order in which variables are added to the model is determined by the order in which
they are listed in the Model: window.
72 CHAPTER 9. LINEAR REGRESSION
R
2
y.12
R
2
y.1
. Indeed, this is the value calculated on page 337 of the text. The
associated p-value of 0.084 means that we would fail to reject the null hypothesis
R
2
y.12
R
2
y.1
= 0 at the .05 level. This is the conclusion reached on page 337 of
the text. Therefore, we are unable to demonstrate that adding x
2
to a model
that contains x
1
adds to R
2
.
To further elucidate, suppose we were working with four predictors, x
1
,
x
2
, x
3
and x
4
and had entered them in that order in the Model: box. The
obtained F and associated p-values for each predictor would represent tests of
the following null hypotheses.
Source Hypothesis
x
1
R
2
y.1
= 0
x
2
R
2
y.12
R
2
y.1
= 0
x
3
R
2
y.123
R
2
y.12
= 0
x
4
R
2
y.1234
R
2
y.123
= 0
Exercises 73
Exercises
9.1 Find the regression equation for predicting y from x for the data in Table
9.1 on page 323.
9.2 Use MINITAB to construct the model described in Exercise 9.7 on page
341 of the text.
9.3 Test the null hypothesis H
0
: R
2
= 0 for the model you constructed in
Exercise 9.2.
9.4 Use the data in Table 9.2 on page 330 to determine whether adding x
1
to a
model that contains x
2
adds signicantly to
R
2
.
74 CHAPTER 9. LINEAR REGRESSION
Chapter 10
Methods Based on the
Permutation Principle
10.1 Introduction
In this chapter you will use MINITAB to perform Wilcoxons signed-ranks test,
Wilcoxons rank-sum (Mann-Whitney) test, the Kruskal-Wallis test and Fishers
exact test.
10.3 Applications (page 348)
Task 10.1
Perform the Wilcoxons signed-ranks test described in Example 10.15 on page
369.
Solution
Open the table 10.7 data.MTW worksheet. Notice that this le contains
dierence scores rather than the pre- and post-treatment values. Select Stat
Nonparametrics1-Sample Wilcoxon.... Type C1 or d in the Variables:
window. Darken the Test median: button and type 0.0 in the accompanying
window if necessary. Be sure the Alternative: window is set to not equal.
Click OK.
MINITAB reports the test p-value as 0.262 which results in failure to reject
the null hypothesis at = .05. This is the same conclusion reached in the text.
It should be noted, however, that p-values reported by MINITAB for Wilcoxons
signed-ranks are approximations rather than exact values.
75
76 CHAPTER 10. PERMUTATION BASED METHODS
Figure 10.1: Data arranged for
Kruskal-Wallis test analysis.
Task 10.2
Perform the Wilcoxons rank-sum test described in Example 10.20 on page 382.
Solution
Enter the data accompanying Example 10.20 into a new worksheet. Select
StatNonparametricsMann-Whitney....
1
The Mann-Whitney page
opens. Type C1 in the First Sample: window and C2 in the Second Sample:
window. Set Alternative: to less than and click OK.
MINITAB gives the one-tailed p-value as 0.3313. As noted in the solution
to Example 10.20, the exact p-value is .35 so that the approximation provided
by MINITAB is quite good even for this case in which sample size is small.
Task 10.3
Perform a Kruskal-Wallis test on the raw data in the table on page 393.
Solution
Enter the group number, i.e. 1, 2, or 3, of each of the 6 subjects in C1 and their
outcome measure in C2. The entries should appear as in Figure 10.1. Now se-
lect StatNonparametricsKruskal-Wallis.... The Kruskal-Wallis page
opens. Enter C2 in the Response: window and C1 in the Factor: window.
Click OK. MINITAB gives the p-value for the test as 0.156. As is shown in
the solution to this problem (page 395), the exact p-value is .20 so that the
MINITAB approximation is not too good. But this is to be expected with such
small sample sizes.
1
Recall that the Mann-Whitney test is just a dierent form of the Wilcoxon rank-sum test.
10.3. APPLICATIONS 77
Task 10.4
Perform a two-tailed Fishers exact test on the data in Table 10.23 on page 405.
Solution
Open the table 10.23 data.MTW worksheet. Notice that C1 contains row
numbers and C2 the column numbers of the cells that makeup Table 10.23.
Taken together, these columns uniquely identify each cell of the table. Column
C3 provides the frequency in each of the four cells.
Select StatTablesCross Tabulation and Chi-Square.... The Cross
Tabulation and Chi-Square page opens. Type C1 in the For rows: window,
C2 in the For columns: window and C3 in the Frequencies are in: window.
Now click Other Stats... and place a check in the Fishers exact test for
2x2 tables box. Click OK twice.
MINITAB gives the p-value for the test as 0.0889111 which is the value
calculated for a two-tailed test in the solution to Example 10.30 on page 405.
Notice that MINITAB did not provide the option of performing a one-tailed
test.
78 CHAPTER 10. PERMUTATION BASED METHODS
Exercises
10.1 Conduct the test described in Example 10.16 on page 371.
10.2 Perform the test alluded to in Example 10.21 on page 383.
10.3 Conduct the test described in Example 10.22 on page 385.
10.4 Conduct the Kruskal-Wallis test described in Example 10.27 on page 395.
10.5 Conduct the test described in Example 10.31 on page 407.
Appendix A
Table Relating Text Items
to Tasks
This table relates various portions of the text to specic tasks and the task page
number.
Text Task Page
Table 2.2 2.1 11
Table 2.3
*
2.2 13
Table 2.4 2.3 13
Figure 2.1 2.4 15
Table 2.3
*
2.5 17
Figure 2.4 2.6 18
Figure 2.6 2.7 19
Example 2.4 2.9 24
Example 2.7 2.10 25
Table 2.6 2.11 26
Example 2.17 2.12 28
Example 2.21 2.13 28
Example 2.23 2.14 28
Example 2.24 2.15 29
Table 3.1 3.1 31
Example 3.9 3.2 31
Table 4.1 4.1 35
Example 4.7 4.2 36
Example 4.9 4.3 36
Example 4.12 4.4 37
79
80 APPENDIX A. TABLE RELATING TEXT ITEMS TO TASKS
Text Task Page
Example 4.14 4.5 38
page 108 4.6 38
Example 4.19 4.7 39
Example 4.25 4.8 39
Example 4.26 4.9 40
Example 4.32 4.10 41
Example 4.34 4.11 41
Example 4.36 4.12 42
Example 4.37 4.13 42
Example 4.41 4.14 43
Example 4.42 4.15 43
Example 5.1 5.1 45
Example 5.4 5.2 46
Table 5.1 5.3 46
Table 5.7 5.4 46
Example 5.13 5.5 48
Example 5.15 5.6 48
Example 5.17 5.7 48
Example 5.20 5.8 49
Example 5.22 5.9 49
Example 5.24 5.10 49
Example 5.27 5.11 50
Example 5.30 5.12 50
Example 6.1 6.1 53
Example 6.3 6.2 54
Example .6.5 6.3 54
Example 6.7 6.4 55
Example 6.9 6.5 55
Example 6.11 6.6 55
Table 7.1 7.1 59
Example 7.6 7.2 60
Example 7.9 7.3 61
Table 8.1 8.1 65
Example 8.7 8.2 66
Example 9.1 9.1 69
Table 9.1
*
9.2 69
Table 9.1
*
9.3 70
Table 9.2
*
9.4 70
Table 9.2
*
9.5 70
81
Text Task Page
Table 9.2
*
9.6 70
Example 10.15 10.1 75
Example 10.20 10.2 76
page 393 10.3 76
Table 10.23 10.4 77
* Appears more than once in list.