CH 17 Statistica
CH 17 Statistica
CHAPTER 17
Analysis of Variance
Multiple-Choice Questions
1.
When doing an analysis of variance, where each population is assumed to have the
same variance, we are more likely to reject the null hypothesis of equality of population
means when there is:
A) high variability among (between) the sample means and high variability around within
the sample means.
B) low variability among (between) the sample means and low variability around within
the sample means.
C) high variability among (between) the sample means and low variability around within
the sample means.
D) low variability among (between) the sample means and high variability around within
the sample means.
ANSWER:
C
2.
3.
In a two-way ANOVA, with one observation per cell, if there are 5 groups and 3 blocks,
then the total number of observations is
A) 15.
B) 13.
C) 11.
D) 8.
ANSWER:
217
Chapter 17
QUESTIONS 4 THROUGH 9 ARE BASED ON THE FOLLOWING INFORMATION:
An analysis of variance table is displayed below with missing values, denoted by asterisks.
ANOVA: Single Factor
SUMMARY
Groups
Column 1
Column 2
Count
7
6
Sum
82
81
Average
13.5
Variance
14.238
4.7
Column 3
112
16
10.667
Column 4
80
13.333
11.467
ANOVA
Source of Variation
Between Groups
Within Groups
SS
230.262
df
3
22
Total
295.884
4.
5.
218
7.
6.
MS
21.874
P-value
F critical
3.0491
Analysis of Variance
8.
9.
South
Midwest
West
Low
Medium
87.3
84.3
73.2
69.4
82.3
83.2
76.0
75.2
High
74.5
72.3
73.4
68.4
Count
4
4
Sum
*A*
Average
79.7
78.025
Variance
40.15333
49.50917
High
72.15
7.056667
East
82.03333
44.81333
South
71.63333
3.943333
Midwest
79.63333
29.34333
West
73.2
*B*
ANOVA
Source of Variation
Rows
Columns
SS
125.765
224.843
Df
2
3
MS
62.8825
74.946
Error
65.315
Total
415.923
F
F
G
P-value
0.039939
0.02274
219
Chapter 17
10.
11.
12.
220
16.
15.
14.
13.
Analysis of Variance
UESTIONS 17 THROUGH 19 ARE BASED ON THE FOLLOWING INFORMATION:
The results of a two-factor ANOVA without replication are displayed below:
East
South
Midwest
West
A
B
739
726
791
748
733
743
779
754
730
780
755
803
Count
4
4
Sum
3042
2971
Average
760.5
742.75
Variance
830.333
144.917
3068
767
992.667
East
2195
731.667
44.333
South
2319
773
499
Midwest
2231
743.667
121.333
West
2336
778.667
600.333
ANOVA
Source of Variation
Rows
Columns
SS
1260.5
4634.25
df
2
3
MS
630.25
1544.75
Error
1269.5
211.583
Total
7164.25
11
17.
P-value
0.1263
0.0199
F critical
5.143249
4.757055
18.
F
2.9787
7.3009
Is there sufficient evidence to reject H 0 that the average between different regions is the
same?
A) Reject H 0 for = 0.01.
221
Chapter 17
19.
Is there sufficient evidence to reject H 0 that the average between programs A, B, and C
regions is the same?
A) Unable to reject H 0 for < 0.10.
B) Reject H 0 for = 0.10.
C) Reject H 0 for = 0.01.
D) Reject H 0 for = 0.05.
ANSWER:
A
21.
222
23.
22.
SS
69.72
30.40
76.64
df
A
B
C
MS
D
E
F
F
G
H
Analysis of Variance
24.
25.
26.
28.
27.
Is there sufficient evidence to reject that there is no difference between the analysts
forecasts?
A) Reject H 0 for = 0.05.
Is there sufficient evidence to reject H 0 that there is no difference between the stocks?
A) Unable to reject H 0 for < 0.10.
B) Reject H 0 for = 0.05.
223
Chapter 17
QUESTIONS 30 THROUGH 39 ARE BASED ON THE FOLLOWING INFORMATION:
A consumer group wanted to examine how long different life bulbs last. They selected three
different light bulbs (60 watt, 75 watt and 100 watt) from four different manufacturers. They
generated the following statistics from the data.
Source of Variation
Between types of bulbs
Between manufacturers
Error
30.
31.
32.
224
34.
33.
SS
1454
739
1193
df
A
B
C
MS
D
E
F
F
G
H
Analysis of Variance
35.
36.
37.
38.
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the different types of light bulbs?
A) Unable to reject H 0 for < 0.10
B) Reject H 0 for = 0.01
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the different manufacturers of light bulbs?
A) Reject H 0 for = 0.01
B) Reject H 0 for = 0.25
C) Reject H 0 for = 0.05
225
Chapter 17
41.
42.
43.
44.
The numerator and denominator degrees of freedom for the F test in a one-way ANOVA
with n observations and K groups are, respectively, given by
A) (n K) and (K 1).
B) (K 1) and (n K).
C) (K n) and (n 1).
D) (n 1) and (K n).
ANSWER:
B
45.
In a one-way ANOVA, the null hypothesis is always written in a way to reflect that
A) there is no treatment effect.
B) there is some treatment effect.
C) all the population means are different.
D) some of the population means are different.
ANSWER:
A
46.
ANSWER:
226
Analysis of Variance
47.
An airline wants to select a computer software package for its reservation system. Four
software packages (1, 2, 3, and 4) are commercially available. The airline will choose the
package that bumps as few passengers, on the average, as possible during a month. An
experiment is set up in which each package is used to make reservations for five
randomly selected weeks. (A total of 20 weeks was included in the experiment.) The
number of passengers bumped each week is given below. How should the data be
analyzed?
Package 1: 12, 14, 9, 11, 16,
Package 2: 2, 4, 7, 3, 1
Package 3: 10, 9, 6, 10, 12,
Package 4: 7, 6, 6, 15, 12
A) F test for differences in variances.
B) One-way ANOVA F test.
C) t test for the differences in means.
D) t test for the mean difference between matched pairs.
ANSWER:
B
48.
SS
df
MS
P-value
Neighborhoods
Error
2.97
1.43
3
12
0.990
0.119
8.31
0.0260
Total
4.40
15
In a Kruskal-Wallis test at the 5% significance level, there are four samples and the value
of the test statistic is calculated as W = 8.79. The most accurate statement that can be
made about the p-value is that it is:
A) greater than 0.10
B) greater than 0.05 but smaller than 0.10
C) greater than 0.05
D) greater than 0.025 but smaller than 0.05
ANSWER:
D
227
Chapter 17
50.
A randomized block design ANOVA has five treatments and four blocks. The computed
test statistic (value of F) is 4.35. With a 0.05 significance level, the appropriate table
value and conclusion will, respectively, be:
A) 14.37; fail to reject the null hypothesis.
B) 3.11; fail to reject the null hypothesis.
C) 3.26; fail to reject the null hypothesis.
D) 3.26; reject the null hypothesis.
ANSWER:
D
51.
A randomized block experiment having five treatments and six blocks produced the
following values: SST = 1446, SSG = 287 and SSE = 180. The value of SSB must be:
A) 30
B) 979
C) 467
D) 20
ANSWER:
52.
Three tennis players, a beginner, an intermediate, and advanced, have been randomly
selected from the membership of a racquet facility club in a large city. Using the same
tennis ball, each player hits ten serves, one with each of three racquet models, with the
three racquet models selected randomly. The speed of each serve is measured with a
machine and the result recorded. Among the ANOVA models listed below, the most likely
model to fit this situation is the:
A) one-way ANOVA
B) two-way ANOVA
C) randomized block design
D) None of the above
ANSWER:
C
53.
A randomized block design with 4 treatments and 5 blocks produced the following sum of
squares values: SST = 1951, SSG = 349, SSB = 1414. The value of SSE must be:
A) 188
B) 537
C) 1602
D) 1763
ANSWER:
228
Analysis of Variance
True-False Questions
54.
When doing an analysis of variance, we are more likely to reject the null hypothesis of
equality of population means when there is low variability between the sample means.
ANSWER:
F
55.
When doing an analysis of variance, we are more likely to reject the null hypothesis of
equality of population means when there is low variability around the sample means.
ANSWER:
T
56.
The Kruskal - Wallis test is used in two-way ANOVA when we have reason to believe the
parent distribution is not normal.
ANSWER:
F
57.
Consider a two-way ANOVA with multiple observations per cell. It is possible to reject the
null hypothesis of no interaction, but be unable to reject the null hypothesis of no effect
arising from either the block variable or the group variable.
ANSWER:
T
58.
59.
The Kruskal-Wallis test is employed to test the equality of population means when an
investigator has strong grounds for suspecting that the parent population distributions
may be markedly different from the normal.
ANSWER:
T
60.
In order to calculate the Kruskal-Wallis test, the values within each sample are ranked in
ascending order, using the average of adjacent ranks in the case of ties.
ANSWER:
F
61.
In one-way ANOVA, if the ratio F =MSG / MSW is quite close to 1.0, there is little cause to
doubt the null hypothesis of equality of population mean.
ANSWER:
T
62.
In one-way ANOVA, if the variability between groups is large compared to the variability
within groups, we suspect the null hypothesis of equality of population means to be false.
ANSWER:
T
63.
The Kruskal-Wallis test procedure is approximately valid, provided that the sample
contains at least ten observations from each population.
ANSWER:
F
64.
In a two-way ANOVA, with one observation per cell, the total sum of squares, SST, can
be decomposed into three components: between blocks sum of squares, SSB, between
groups sum of squares, SSG, and error sum of squares, SSE.
ANSWER:
T
65.
In a two-way ANOVA, with one observation per cell, if there are 4 groups and 3 blocks,
then there is a total of 7 observations.
ANSWER:
F
229
Chapter 17
66.
A study will be undertaken to examine the effect of two kinds of background music and of
two assembly methods on the output of workers at a fitness shoe factory. Two workers
will be randomly assigned to each of four groups, for a total of eight in the study. Each
worker will be given a headphone set so that the music type can be controlled. The
number of shoes completed by each worker will be recorded. Does the kind of music or
the assembly method or a combination of music and method affect output? The ANOVA
model most likely to fit this situation is the Spearman rank correlation model.
ANSWER:
F
67.
In a two-way ANOVA, a design with only one observation per cell allows the isolation of a
further source of variability the interaction between groups and bocks.
ANSWER:
F
68.
In a two-way ANOVA, interaction between groups and blocks occur when differences in
group efforts are not distributed uniformly across blocks.
ANSWER:
T
69.
70.
In a two-way ANOVA, with more than one observation per cell, the total sum of squares,
SST, can be decomposed into four components: between groups sum of squares, SSG,
between blocks sum of squares, SSB, interaction sum of squares, SSI, and error sum
of squares, SSE.
ANSWER:
T
71.
In a two-way ANOVA, with one or more than one observation per cell, there are three
tests of hypothesis that can be carried out with reference to the F-distribution with the
corresponding numerator and denominator degrees of freedom.
ANSWER:
F
72.
The analysis of variance (ANOVA) tests hypotheses about the population variance.
ANSWER:
F
73.
The F test in one-way ANOVA model is just an expansion of the t test for independent
samples.
ANSWER:
T
74.
75.
Subject to the value of the sample sizes, the MSE can be negative (especially for very
small sample sizes) or positive.
ANSWER:
F
230
Analysis of Variance
76.
A study will be undertaken to examine the effect of two kinds of background music and of
two assembly methods on the output of workers at a fitness shoe factory. Two workers
will be randomly assigned to each of four groups, for a total of eight in the study. Each
worker will be given a headphone set so that the music type can be controlled. The
number of shoes completed by each worker will be recorded. Does the kind of music or
the assembly method or a combination of music and method affect output? The ANOVA
model most likely to fit this situation is the one-way analysis of variance.
ANSWER:
F
77.
When the F test is used for ANOVA, the rejection region is always in the right tail.
ANSWER:
T
78.
The critical value of the Kruskal-Wallis test for comparing more than two independent
samples is obtained from the chi-square distribution whenever each sample size is at
least 5.
ANSWER:
T
79.
80.
81.
82.
83.
In the one-way ANOVA test, if the sample sizes are large, then the assumption that the
underlying population distributions are normal is not necessary.
ANSWER:
F
84.
When there are two treatment groups, one-way ANOVA is equivalent to the pooledvariances t-test for difference between means.
ANSWER:
T
85.
A study will be undertaken to examine the effect of two kinds of background music and of
two assembly methods on the output of workers at a fitness shoe factory. Two workers
will be randomly assigned to each of four groups, for a total of eight in the study. Each
worker will be given a headphone set so that the music type can be controlled. The
number of shoes completed by each worker will be recorded. Does the kind of music or
the assembly method or a combination of music and method affect output? The ANOVA
model most likely to fit this situation is the simple regression model.
ANSWER:
F
86.
In ANOVA, if the calculated F-statistic exceeds the critical F for a given test, the null
hypothesis of equal population means is rejected.
ANSWER:
T
231
Chapter 17
87.
If we simultaneously examine the effects of two factors on the dependent variable, along
with the effects of interactions between the different levels of those factors, we are
performing two-way ANOVA.
ANSWER:
T
88.
Basic to ANOVA is the comparison of variation between samples versus the amount of
variation within the samples. The test statistic is an F-ratio in which the numerator
reflects variation within the samples and the denominator reflects the variation between
them.
ANSWER:
F
89.
Like the majority of the nonparametric tests, the Kruskal-Wallis test is based the ranks of
the sample observations.
ANSWER:
T
90.
A study will be undertaken to examine the effect of two kinds of background music and of
two assembly methods on the output of workers at a fitness shoe factory. Two workers
will be randomly assigned to each of four groups, for a total of eight in the study. Each
worker will be given a headphone set so that the music type can be controlled. The
number of shoes completed by each worker will be recorded. Does the kind of music or
the assembly method or a combination of music and method affect output? The ANOVA
model most likely to fit this situation is the two-way analysis of variance.
ANSWER:
T
232
Analysis of Variance
Pennkamp
40
35
33
California
21
21
27
Mayaguez
37
47
45
Maui
22
19
25
Analysis of Variance
Source of Variation
Location
Error
SS
df
MS
P-value
963.6
210.0
4
10
240.9
21.0
11.47
0.001
Total
91.
H 0 : 1 2 3 4 5
What is the value of the element in the ANOVA table that always provides an estimate of
the population variance?
ANSWER:
MSE = 21.0
93.
The test statistic F is the ratio of two elements of the ANOVA table. Of these elements,
what is the value of the one that provides an estimate of the population variance only
when the null hypothesis is true?
ANSWER:
MSG = 240.9
94.
95.
233
Chapter 17
96.
97.
98.
What are the numerator and denominator degrees of freedom of the test ratio in testing
the hypotheses in Question 50?
ANSWER:
The numerator and denominator degrees of freedom are 4 and 10, respectively.
Sum of Squares
1280
780
2060
Degrees of Freedom
4
15
19
H 0 : 1 2 3 4 5
101.
102.
103.
234
Analysis of Variance
QUESTIONS 104 THROUGH 106 ARE BASED ON THE FOLLOWING INFORMATION:
A machine press manufacturer is comparing orders coming from different regions in the country
and has collected the following data from 20 orders.
East
South
Midwest
West
7
9
5
6
2
4
5
7
10
4
104.
Groups
Column 1
Column 2
Count
4
6
Column 3
25
Column 4
34
6.8
2.2
Source of Variation
SS
Between Groups
24.5667
Within Groups
50.3833
df
3
16
MS
8.1889
3.1490
F
2.6005
Total
19
ANOVA
105.
74.95
P-value
0.088
Is there sufficient evidence to reject the null hypothesis that the average number of
machines per order is the same for different regions of the country?
ANSWER:
H 0 : 1 2 3 4
235
Chapter 17
106.
Carry out the Kruskal-Wallis test for this data using statistical software and write your
conclusion?
ANSWER:
Since p-value = 0.0949, we fail to reject H 0 for all levels of significance greater than
9.49%, and conclude that there is sufficient statistical evidence that the average number
of machines per order is the same for the four different regions of the country.
QUESTIONS 107 THROUGH 109 ARE BASED ON THE FOLLOWING INFORMATION:
A marketing manager is considering adopting one of three new marketing schemes. He chooses
three sales people in each region, and has them try the new schemes. He then records the
resulting sales as shown below.
Marketing
Schemes
A
B
C
236
East
South
Midwest
West
4
7
5
6
7
8
4
9
Analysis of Variance
107.
Count
4
4
Sum
20
30
Average Variance
5
2
7.5
1.6667
27
6.75
2.9167
East
17
5.6667
2.3333
South
18
Midwest
24
West
18
SS
13.1667
10.25
df
2
3
MS
6.5834
3.4167
F
4.1580
2.1580
Error
9.5
1.5833
Total
32.9167
11
ANOVA
Source of Variation
Rows
Columns
108.
P-value
0.0736
0.1942
Test at the 10% level of significance the null hypothesis that the population mean sales
are the same for the four regions.
ANSWER:
H 0 : 1 2 3 4
Since p-value = 0.1942, we fail to reject H 0 at the 10% level of significance, and
conclude that the population mean sales are the same for the four regions.
109.
Test at the 10% level of significance the null hypothesis that the population mean sales
are the same for the three marketing schemes.
ANSWER:
H 0 : 1 2 3
Since p-value = 0.0736, we reject H 0 at the 10% level of significance, and conclude that
there is difference in sales for at lease two of the three marketing schemes.
237
Chapter 17
QUESTIONS 110 THROUGH 112 ARE BASED ON THE FOLLOWING INFORMATION:
Five stock analysts were asked to predict earnings for four different stocks for the next year. The
following statistics summarize their responses:
Source of Variation
Between Analysts
Between Stocks
Error
Total
110.
Sum of Squares
74.6
35.7
82.2
192.5
111.
SS
74.6
35.7
82.2
192.5
df
4
3
12
19
MS
18.65
11.90
6.85
F-ratio
2.723
1.737
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the analysts forecast?
ANSWER:
Since F4, 12, 0.05 = 3.26 and F =2.723, we fail to reject H 0 at 0.05 . There is no
sufficient statistical evidence to conclude that there is a difference between the analysts
forecast.
112.
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the stocks?
ANSWER:
Since F3, 12, 0.05 =3.49 and F =1.737 we fail to reject H 0 0.05 . There is no sufficient
statistical evidence to conclude that there is a difference between the stocks.
Sum of Squares
174.50
258.75
302.40
238
SS
174.50
258.75
302.40
df
2
9
18
29
MS
87.25
28.75
16.8
F-ratio
5.193
1.711
Analysis of Variance
114.
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the three exam forms? Use = 0.05
ANSWER:
Since F2, 18, 0.05 = 3.55 and F = 5.193, we reject H 0 at = 0.05 and conclude that there
is sufficient statistical evidence that there a difference between at least two of the three
exam forms.
115.
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the applicants? Use = 0.05
ANSWER:
Since F9, 18, 0.05 = 2.46 and F =1.711, we fail to reject H 0 0.05 , and conclude that
there is no difference between the applicants.
Sum of Squares
240.3
122.5
20.3
325.3
117.
SS
240.3
122.5
20.3
325.3
708.4
df
3
4
12
60
79
MS
80.1
30.625
1.692
5.422
F-ratio
14.773
5.648
0.312
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the brokerage houses? Use = 0.05
ANSWER:
Since F3, 60, 0.05 =2.76 and F =14.773, we reject H 0 .There is sufficient statistical evidence
to conclude that there is a difference between at least two of the four brokerage houses.
239
Chapter 17
118.
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the stocks earning forecasts? Use = 0.05
ANSWER:
Since F4, 60, 0.05 =2.53 and F = 5.648, we reject H 0 at = 0.05. There is sufficient
statistical evidence to conclude that there is a difference between at least two of the five
stocks earning forecasts.
119.
Is there sufficient evidence to reject the null hypothesis that there is no interaction
between brokerage houses and stocks earning forecasts? Use = 0.05
ANSWER:
Since F12, 60, 0.05 =1.92 and F = 0.312, we fail to reject H 0 at 0.05 , and conclude that
there is no interaction between the brokerage houses and the stocks earnings forecasts.
Sum of Squares
36.40
13.40
31.32
125.2
121.
SS
36.40
13.40
31.32
125.2
206.32
df
3
4
12
60
79
MS
12.133
3.35
2.61
2.087
F-ratio
5.814
1.605
1.251
Is there sufficient evidence to reject the null hypothesis that there is no difference
between the regions of the country? Use = 0.05
ANSWER:
Since F3, 60, 0.05 =2.76 and F = 5.814, we reject H 0 . There is sufficient statistical evidence
to conclude that differences exist in at least two regions of the country.
122.
Is there sufficient evidence to reject the null hypothesis that there is no difference
between different types of brochures? Use = 0.05
ANSWER:
240
Analysis of Variance
Since F4, 60, 0.05 =2.53 and F =1.605, we fail to reject H 0 at 0.05 , and conclude that
there is no difference between the different types of brochures.
123.
Is there sufficient evidence to reject the null hypothesis that there is no interaction
between region of the country and brochure type? Use = 0.05
ANSWER:
Since F12, 60, 0.05 =1.92 and F =1.251, we fail to reject H 0 at 0.05 , and conclude that
there is no interaction between regions of the country and brochure type.
Sum of Squares
1,600
*
2790
Degrees of Freedom
*
15
17
125.
126.
H 0 : 1 2 3
128.
129.
130.
241
Chapter 17
ANSWER:
Reject H 0 if F F2,15,0.05 3.68. Therefore, H 0 is rejected. We conclude that at least two
of the population means are not equal.
QUESTIONS 131 THROUGH 135 ARE BASED ON THE FOLLOWING INFORMATION:
A production manager monitors the output per hour of four workers over the course of four hours.
The results are reproduced below.
Worker A
Worker B
Worker C
Worker D
7
3
5
3
5
8
6
2
Assume that the populations from which the sample drawn are normally distributed.
131.
Count Sum
4
22
4
14
Average Variance
5.5
3
3.5
1.6667
Worker C
28
Worker D
16
3.3333
Source of Variation
Between Groups
Within Groups
SS
30
30
df
3
12
MS
10
2.5
F
4
Total
60
15
ANOVA
132.
P-value
0.0346
What are the appropriate null and alternative hypothesis using this information?
ANSWER:
H 0 : 1 2 3 4
242
Assume that the populations from which the samples are drawn are normally distributed,
is there sufficient evidence to reject the H 0 that there is no difference in the output per
hour of the four workers?
Analysis of Variance
ANSWER:
Since p-value = 0.0346, we reject H 0 at =0.05 and conclude that there is sufficient
statistical evidence that there is a difference in the output per hour for at least one pair of
workers (possibly workers B and C or C and D).
134.
Carry out the Kruskal-Wallis test for this data at =0.05 and write your conclusion.
ANSWER:
Compare your answers to Questions 133 and 134 and explain how to interpret these
results.
ANSWER:
In Question 133 we rejected H 0 at = 0.05, but failed to reject H 0 under same level of
significance in Question 134. The reason is that under the assumption of normality in the
populations, nonparametric tests such as Kruskal-Wallis are less powerful than
parametric tests, such as ANOVA.
243
Chapter 17
ANSWER:
The null hypothesis tests the equality of the population mean ratings across the classes;
that is, H 0 : 1 2 3 .
137.
Carry out the appropriate test for the null hypothesis in Question 136 at the 10% level.
ANSWER:
W = 0.75
2
2,0.10
4.61 , therefore, do not reject the null hypothesis at the 10% level. We conclude
that the population mean ratings across the classes (sophomores, sophomores, and
seniors) are the same.
QUESTIONS 138 THROUGH 140 ARE BASED ON THE FOLLOWING INFORMATION:
Random samples of eight freshmen, eight sophomores, and eight juniors taking a business
statistics class were drawn. The accompanying table shows scores on the final examination.
Freshmen
81
73
56
77
72
96
64
85
138.
Sophomores
80
74
69
97
81
88
65
74
Juniors
90
81
90
94
59
76
78
67
139.
Test the null hypothesis that the three population mean scores are equal. Use = 0.05
ANSWER:
H 0 : 1 2 3
244
Analysis of Variance
Since p-value = 0.7862, we fail to reject H 0 at = 0.05, and conclude that the three
population mean scores are equal
140.
Carry out a nonparametric test of the null hypothesis of equality of population mean
examination scores for freshman, sophomore, and juniors. Use = 0.05
ANSWER:
We use the Kruskal-Wallis test as the nonparametric alternative to the one-way ANOVA
test. The results are shown below
Consider a problem with three subgroups with the sum of ranks in each of the subgroups
equal to 32, 101.5, and 76.5, with subgroup sizes equal to 7, 7, and 6, respectively.
Complete the Kruskal-Wallis test and test the null hypothesis of equal subgroup ranks at
the 1% level.
ANSWER:
H 0 : 1 2 3
Ri2
12
12 322 101.52 76.52
3(n 1)
3(21) = 11.097
n(n 1) i ni
20(21) 7
7
6
245
Chapter 17
Reject H 0 if W k 1,0.01 2,0.01 = 9.21. Therefore H 0 is rejected at the 1% kevel. We
conclude that at least two of the three population means are not the same.
2
Associate
14
9
16
13
16
Full
9
8
10
7
8
12
143.
Count
7
5
6
Sum
86
68
54
SS
63.8159
94.6286
158.4445
df
2
15
17
Average Variance
12.28571 7.571429
13.6
8.3
9
3.2
MS
31.9079
6.3086
F
5.0579
P-value
0.0209
F crit
3.6823
Test the null hypothesis that the three population mean times are equal. Use = 0.05
ANSWER:
H 0 : 1 2 3
Since p-value = 0.0209, we reject H 0 at = 0.05, and conclude that at least two of the
three population mean times are not the same.
144.
Without assuming normal population distributions, test the null hypothesis that the
population mean times spent outside the classroom on teaching responsibilities are the
same for assistant, associate, and full professors.
ANSWER:
246
Analysis of Variance
Without assuming normal population distributions, we will employ Kruskal-Wallis test as
the nonparametric alternative to the one-way ANOVA test. The results are shown below.
Again, Since p-value = 0.0298, we reject H 0 at = 0.05, and conclude that at least two
of the three population mean times are not the same.
QUESTIONS 145 THROUGH 150 ARE BASED ON THE FOLLOWING INFORMATION:
Consider a two-way analysis of variance with one observation per cell and randomized blocks
with the following results:
Source of Variation
Between groups
Between blocks
Error
Total
145.
Sum of Squares
456.0
278.4
464.4
1198.8
Degrees of Freedom
5
5
25
35
146.
147.
247
Chapter 17
148.
149.
SS
456.0
278.4
464.4
1198.8
df
5
5
25
35
MS
91.2
55.68
18.576
F-ratio
4.1
3.0
Test the hypotheses that between-group means are equal. Use = 0.05
ANSWER:
H 0 : 1 2 3 4 5 6
H 0 : 1 2 3 4 5 6
What is the appropriate statistical technique that can be used to analyze such data?
ANSWER:
The Kruskal - Wallis test
248
Analysis of Variance
152.
Apply the technique in Question 158 and test to determine if there is enough evidence at
the 5% significance level to infer that at least one of the population means differs from the
others.
ANSWER:
H 0 : 1 2 3
H1 : At least one mean differs from the others
2
Reject H 0 if W 2,0.05 5.991
Test statistic: W = 0.38; therefore, we fail to reject the null hypothesis. We may infer that
the population means are equal.
153.
A
5.2
4.9
Text
B
4.7
4.5
C
4.8
4.2
249
Chapter 17
Mix
154.
5.0
4.5
4.7
155.
Test the null hypothesis of equality of population mean ratings for the three texts. Use
= 0.05
ANSWER:
H 0 : 1 2 3
Since p-value = 0.023, H 0 is rejected at = 0.05. We conclude that at least two of the
population mean ratings for the three texts are not the same.
156.
Test the null hypothesis of equality of population mean ratings for the three examination
types. Use = 0.05
ANSWER:
H 0 : 1 2 3
Since p-value = 0.0772, H 0 is not rejected at = 0.05. We conclude that the population
mean ratings for the three examination types are the same.
250
Analysis of Variance
A
7
6
5
7
B
5
8
4
4
8
5
7
6
C
6
5
6
8
9
7
6
7
D
8
8
7
5
9
8
7
6
9
9
8
7
158.
Test the null hypothesis that the population mean ratings are the same for the four
dormitories.
ANSWER:
H 0 : 1 2 3 4
Test the null hypothesis that the population mean ratings are the same for the four
student years.
ANSWER:
H 0 : 1 2 3 4
251
Chapter 17
160.
Test the null hypothesis of no interaction between student year and dormitory rating.
ANSWER:
H 0 : No interaction exists between year and dormitory ratings
252