Exercises 513c. How does the answer to part (b) relate to the answer to part (a)?
d. Plot scatterplots of the relationship between Pretest and Posttest separately for each
group. What do these plots show?
e. Run a test on the null hypothesis that the Gain for the Control is 0.00. What does this
analysis tell you? Are you surprised?
f. Why would significant gains in the two experimental groups not be interpretable with-
out the control group?Discussion Questions
14.26 In Exercise 14.24 we ignored the fact that we have pairs of subjects from the same family.
a. What is wrong with doing this?
b. Under what conditions would it be acceptable to ignore this problem?
c. What alternative analyses would you suggest?
14.27 In Exercise 14.24 you probably noticed that many observations at Time 2 are missing. (This
is partly because for many patients it had not yet been 3 months since the diagnosis.)
a. Compare the means at Time 1 for those subjects who did, and who did not, have data at
Time 2.
b. If there are differences in (a), what would this suggest to you about the data?
In a study of behavior problems in children we asked 3 “judges” to rate each of 20 children
on the level of aggressive behavior. These judges were the child’s Parent, the child’s
Teacher, and the child him/herself (Self). The data follow.
Child 1 2 3 4 5 6 7 8 910111213 14 1516 17181920
Parent 10 12 14 8 16 21 10 15 18 6 22 14 19 22 11 14 18 25 22 7
Teacher 8 13 17 10 18 24 9 16 18 8 24 19 15 20 10 18 19 30 20 10
Self 12 17 16 15 24 24 13 17 21 13 29 23 16 20 15 17 21 25 25 14
These data are somewhat different from the data we saw in Section 14.10 because in that
case the same people judged each child, whereas here the Parent and Self obviously change
from child to child. We will ignore that for the moment and simply act as if we could some-
how have the same parent and the same “self ” do all the ratings.
14.28 What is the reliability of this data set in terms of the intraclass correlation coefficient?
14.29 What do your calculations tell you about the sources of variability in this data set?
14.30Suppose that you had no concern about the fact that one source systematically rates chil-
dren higher or lower than another source. How might you evaluate reliability differently?
14.31 Under what conditions might you not be interested in differences among judges?
14.32 What do you think is the importance of the fact that the “parent” who supplies the parent
rating changes from child to child?
14.33 Strayer, Drews, and Crouch (2006) (which we saw as a between-subjects design in Exercise
11.32) examined the effects of cell phone use on driving ability. They had 40 drivers drive
while speaking on a cell phone, drive while at the legal limit for alcohol (0.08%), and drive
under normal conditions. (The conditions were counterbalanced across drivers.) The data
for this study are found at http://www.uvm.edu/~dhowell/methods7/DataFiles/Ex14–34. Their
hypothesis, based on the research of others, was that driving while speaking on a cell phone
would have as much of an effect as driving while intoxicated. The dependent variable in this
example is “braking reaction time.” The data have exactly the same means and standard
deviations as they found.
a. Run the analysis of variance for a repeated measures design.
b. Use the appropriate contrasts to compare the three conditions. Did the results support
the experimenters’ predictions?