Professional Documents
Culture Documents
Page 1 of 9
1 (a)
Let C represent the mean skill level for children taught by teacher C. Based on
the data, what is the value of this parameters estimate?
[3 pts]
105.14 (the computed mean for group C, shown in the ANOVA output)
1 (b)
Based on the output for scenario 1, use the output as appropriate to determine
whether there any evidence of violation of the normality assumption, and
complete the following statement:
[9 pts]
the p-value on the probability plot
Since
.285
is / is not
therefore there
is / is not
[circle one]
less than
.05
[circle one]
Version 1 KEY
1 (c)
Page 2 of 9
Based on the output for scenario 1, use the output as appropriate to determine
whether there any evidence of violation of the equal spread assumption, and
complete the following statement:
[9 pts]
Since
largest s
smallest s
28.91
21.94
1.3
is / is not
[circle one]
less than
.2.
[circle one]
1 (d)
[6 pts]
H0: A = B = C
H A : The s are not all equal.
[6 pts]
2463/2
1231.5
=
= 1.73
72564/102 711.41
[6 pts]
Version 1 KEY
1 (e)
What percent of the variability in Skill can be accounted for by the Teacher
effect alone (without taking into account any other variables)? [Show any
calculation you perform.]
[4 pts]
R2 =
1 (g)
Page 3 of 9
SS Teacher
2463
=
= .033 or 3.3%
SS Total
75028
In order to help detect a Teacher effect (if it exists), the researchers decided to
include the covariate Age of the student (in months) and to use ANCOVA. Was
this approach successful in detecting a significant Teacher effect? Say yes or no,
and support your answer with the appropriate numerical value.
[3 pts]
Yes .
In the ANCOVA output (on p. 3 of the packet), the teacher p-value is .011
1 (h)
ANCOVA also adjusts the means of the response (Skill) in light of the initial
differences in the covariate (Age).
Find the adjusted mean of Skill for the group taught by teacher C. Show your
calculation.
[6 pts]
Cs raw mean
mean of
covariate
(p. 2 of output)
Cs
slope between
covariate
Y and covariate
average
(p. 3 of output)
(p. 2 of output)
Version 1 KEY
Page 4 of 9
To better understand the Teacher effect on Skill, look at the pairwise comparisons.
1 (i)
[6
pts]
( 2.394, 28.97)
We are 95% confident that the mean score with teacher C is between 2.39 and
28.97 higher than the mean score with teacher B.
Remark: The confidence interval is just lifted directly from the output.
If you tried to duplicate it, you probably wouldnt get the same thing, because
(a) Minitab uses the corrected means, and (b) Minitab adjusts the confidence
percentage to account for multiple comparisons.
1 (j)
[3 pts]
Version 1 KEY
Page 5 of 9
In the experiment, subjects who were members of a diving team were asked to study a list
of words either on land or under water (wearing SCUBA gear) and then were given a
recall test (which could either be administered on land or under water). So the
experiment manipulated the environment in which the subjects studied, as well as
manipulating the environment in which the subjects were tested. The measured response
was the number of words recalled.
There are therefore two factors in the experiment:
Study: Where the subject studied the words [Dry (on land) or Wet (under water)]
Test: Where the subject was tested [Dry (on land) or Wet (under water)]
The response is Score: The number of words the subject recalled.
20 subjects were randomly assigned to each combination of the levels of the two factors
(4 combinations, total of 80 subjects).
Refer to the output on page 4 of the output packet
to answer the following questions
2 (a)
[4 pts]
2 (b)
Version 1 KEY
Page 6 of 9
For each of the following different ways the study could have been designed, fill
in the blanks with the appropriate numbers, assuming we still want to have 20
subjects in each of the 2x2 treatment groups.
[Note, one of these is the actual way the study was designed.]
[3 pts]
[3 pts]
Mixed ANOVA design, with Study repeated, but with Test not repeated:
[3 pts]
Mixed ANOVA design, with Test repeated, but with Study not repeated:
[3 pts]
2 (b)
Version 1 KEY
Page 7 of 9
Do the data provide evidence that the Test effect is significantly dependent on the
Study effect? Say yes or no, and support your answer with the appropriate
numerical value from the relevant part of the output.
[4 pts]
Yes . The p-value for the interaction is 0.000
2 (c)
The relevant means come from the wet column on p. 4 of the output:
(11.67.95) 2(2.674)
1
20
+ 201
= 3.65 1.69
= ( 1.96, 5.35), which is significantly higher than zero
Version 1 KEY
Page 8 of 9
3 (a)
Which of the following is the best description of the analysis appropriate for the
study?
[4 pts]
(i) Two-factor repeated-measures ANOVA.
(ii) Regular (non-repeated-measures) ANOVA.
(iii) Mixed ANOVA design, with Type repeated, but with Size not repeated.
(iv) Mixed ANOVA design, with Size repeated, but with Type not repeated.
3 (b)
Which of the four analyses (I, II, III, or IV) shown on page 5 of the output packet
is the correct analysis for this study?
[4 pts]
Output I
3 (c)
Version 1 KEY
Page 9 of 9
Based on the appropriate analysis that you should have chosen in (b), report on
the significance of the results.
You can phrase your answer like:
There is (is not) significant ......., (F = ....., p-value=.......)
[5 pts]
[End of Exam 2]