Chapter 21—Meta Analysis
I have placed these questions under Chapter 21 because they cut across multiple topics, but they
are not quite like the questions I used for Chapter 21 in the text. The idea is that here we are asking
students to pull together multiple lines of information in addressing a hypothetical study that
someone might meaningfully run.
You may feel that some of the questions for a specific study hint at the answers to other questions.
In those places I would simply omit the question in which you have less interest or invent a new
question to take its place.
MULTIPLE CHOICE QUESTIONS
In an experiment aimed at evaluating the effect of a memory-enhancing drug on the recall of a
previously learned response. Either the drug or a placebo was administered to different groups of
rats before the memory test. There were three trials of the memory test, and the dependent
variable was the time it took the animal to make the correct response.
21.1 The proper analysis to evaluate the drug enhancement of memory is a(n)
21.2 Suppose the analysis reveals a significant drug by trial interaction. What would we
conclude about the effect of the drug on memory?
21.3 What assumption(s) would we need to make for this analysis?
21.4 We want to compute an effect size estimate for the difference between drug and placebo
treatments. We will take as our dependent variable for this computation each animal’s
mean response time over the three trials. What would be the best estimate of effect size?
Test Bank
396
21.5 It seems reasonable that there is some sort of relationship between the number of trials it
took to learn a response and the speed of response on the test trials. It would be
interesting to know how the drug affects this relationship. How might you go about
looking at this question?
21.6 Why do we not need to use a multiple comparison procedure to compare the drug and
control conditions?
The Morris water maze is used to examine spatial learning in animals. Mice, who are good
swimmers, are placed in a large pool of water and must learn to find a clear platform located just
under the water surface. The platform is their only escape from the pool. If they do not find the
platform, they have to continue swimming until we take them out.
21.7 For one of the trials, the platform was removed from the pool and the amount of time the
mouse spent in each of the four quadrants of the tank was recorded. What statistical test
should be conducted on these data?
21.8 We want to compare the amount of time the mice spent swimming in the first trial to the
amount of time they spent swimming in the tenth trial. What statistical test should be
conducted on these data?
21.9 We want to report an effect size estimate that reflects the fact that the animals were able
to find the platform faster on the 10th trial that they did on the first trial. What would be
the best measure to report?
Chapter 21
21.10 When mice perform in a Morris Water maze they sometimes become confused on a
particular trial and take a long time to get to the platform. On other trials they are just
lucky and go straight to it. How can we eliminate or reduce the influence of these unusual
times?
21.11 If we think that we may have some aberrant times, what is the best way to identify them?
A researcher is interested in the social behavior of mice. He hypothesizes that mice from Strain
A will be more social than mice from Strain B, and Strain B will be significantly more social
than mice from Strain C. Mice from each of the three different strains (20 mice in each strain)
are raised in a laboratory. The mice are group housed with members of its own strain. All the
mice receive the exact same amount of food, water, and light over the first 10 weeks of life. At
11 weeks of age, each mouse is placed in a cage with a novel mouse from his own strain. The
mouse’s investigatory behavior toward the other mouse is timed for 10 minutes.
21.12 Which statistical analysis is most appropriate for testing this researcher’s hypothesis?
21.13 If, after running the appropriate analysis, the researcher finds a result with p = .50, what
can she conclude?
21.14 The researcher then decided to run the experiment again. This time mice of each strain
were singly housed, isolated from members of their own strain for the 11 weeks. Now
she finds that p = .048. This suggests
21.15 Given the above result, what might the experimenter do next?
Test Bank
21.16 If the experimenter does decide to run this experiment again, what might the
experimenter do to improve the chances of finding meaningful results?
A researcher is interested in studying changes in marital satisfaction over the course of marriage.
She hypothesizes that marital satisfaction is highest when couples are newly married, that it
decreases during the middle years of marriage, but rises again during later years of marriage. In
order to test her hypothesis, she recruits fifty married couples within each of the following
conditions: (1) married less than 10 years; (2) married 15-20 years; (3) married more than 25
years. All 150 couples (50 from each condition) complete a standardized measure of marital
satisfaction.
21.17 What is the independent variable?
21.18 What is the dependent variable?
21.19 Which statistical analysis is most appropriate for testing this researcher’s hypothesis?
21.20 Which statistic would result from the appropriate analysis?
21.21 If there is a significant change with length of marriage, the researcher would find
Chapter 21
21.22 Why could the researcher not simply correlate the marital satisfaction score with the
actual years of marriage?
21.23 What assumptions would the researcher have to make to analyze these data correctly?
21.24 If the experimenter had reason to believe that the data would be better analyzed by a
nonparametric test, which would be the most appropriate test?
Many studies have been conducted on the effects of having a pet on the response to medical
problems and on mental health. But doctors in Berlin were interested in the effects of having a
pet bird on lung cancer. (I assume they were wondering if birds might cause air pollution that
could lead to lung cancer.) They asked 239 patients with lung cancer and 429 controls without
lung cancer whether they had a pet bird. Ninety eight of the 239 patients did have a bird, while
101 of the 429 controls had a bird. (That strikes me as a very high percentage of bird owners, but
those are the data.)
21.25 What would be the appropriate analysis to ask if bird ownership was significantly
associated with lung cancer?
21.26 We really do need a measure of effect size in this example. What would be an appropriate
one?
21.27 The odds of having cancer for the bird owners are
Test Bank
400
21.28 Why would it be wrong to calculate relative risk in this study?
21.29 What assumptions would need to underlie our analysis for this experiment?
21.30 Without actually doing the arithmetic on paper, what conclusions would you be likely to
draw from this study?
21.31 How might this study be “improved” as a conclusive test of the hypothesis that birds
cause cancer?
An instructor in Introductory Psychology wanted to know if the textbook she was using was as
effective as other textbooks on the market. In order to test this, she decided that she would teach
the course in the fall with the text she typically uses, and in the spring she would use a new text.
After doing this, she compared the students’ final grades and students’ responses to an ‘end of
semester’ evaluation question. The question was “How effective was the text for this course?”
She discovered that students’ grades increased from a mean of 78 in the fall to 83 in the spring.
She also discovered that the students’ opinion of the course’s text increased from a mean of 4.5
(on a 7-point scale, the higher the better) in the fall to 4.9 in the spring. After running the
appropriate statistical test(s), she discovered that both of these differences were statistically
significant.
21.32 Which of the following tests would have been the most appropriate for this design?
21.33 What is the dependent variable in this study?
Chapter 21
401
21.34 How might this study be improved?
21.35 How can we tell if the difference in the student opinion in the two semesters was a
meaningful difference?
21.36 If we wanted to compute a standardized mean difference (
ˆ
d
) in student satisfaction, our
best standard deviation to use in the calculation could be
21.37 We have the most persuasive evidence of some phenomenon when
21.38 If we see a relationship over time between television violence and children’s expression
of aggression, we could conclude that
TRUE/FALSE QUESTIONS
variables, you will most likely use Pearson correlation.
Test Bank
402
21.40 [FALSE] When you are interested in the degree of relationship between 1 continuous
dependent variable and multiple continuous predictor variables you will most likely use a
Spearman correlation.
or more independent groups.
continuous.
underlying parametric tests may be violated.
being categorized.
OPEN-ENDED QUESTIONS
For each of the following examples, a) specify the independent and dependent variable b) note if
they are categorical or continuous, c) note the type of question, and d) indicate which analysis
would be most appropriate.
21.49 Mary wants to know if there are treatment effects on acting out behavior (the number of
incidents per month) between a group of adolescents who are in a wait list control group
and a group of adolescents receiving Cognitive Behavioral Therapy (CBT).
21.50 Mary wants to know if an equal number of males and females volunteer for each of her
groups (control group and CBT group).
21.51 Mary’s advisor recommended that she add a third group, who are in a support only
condition to her design.
21.52 Another member of Mary’s committee recommended that she change this to a factorial
design. That is, some people should receive CBT only, some support group only, some
both, and some should be a waitlist control. [Everyone on a dissertation committee thinks
that the study should be redesigned to address the question in which they are most
interested. ☺]
Chapter 21
403
21.53 Mary’s advisor is involved in a project examining the frequency of acting out in a group
of 100 adolescents who she is contacting at age 12, 13, 14, and 15 to examine trends in
acting out over time.
21.54 John is interested in studying delinquent behavior. He thinks that low parental
monitoring will be associated with more delinquent behaviors among adolescents. He
plans to measure monitoring using a self-report measure of how often parents ask their
children where they are going, with who, and how often parents do not know where their
children are, etc. The measure of delinquent behavior is the adolescents self-report of the
number of times they have engaged in delinquent acts (e.g., shoplifting, driving under the
influence, fighting, graffiti, etc) during the past year.
21.55 John’s advisor suggests that John consider other factors, besides parental monitoring,
such as the number of delinquent friends and the strength of future career aspirations that
may contribute to delinquency.
21.56 An undergraduate working with John is doing an honors thesis using some of the data
that John collected. She wants to know if there is a difference in the amount of parental
monitoring for mothers versus fathers of the same child.
21.57 As part of a small side project, John wants to know if fathers monitor daughters more
than sons.
Answers to Open-ended Questions
Chapter 21.
Test Bank
404