Answers to
Study Questions
for
Chapter 11
(Don’t forget that the companion website also has multiple choice
questions that you can take for practice. You will find them here: http://www.southalabama.edu/coe/bset/johnson/dr_johnson/2mcq.htm)
As a starting point, let’s
quickly review the three necessary conditions for establishing cause and
effect:
Condition 1: Variable A and variable B must be related (the relationship
condition).
Condition 2: Proper time order must be established (the temporal antecedence
condition).
Condition 3: The relationship between variable A and variable B must not
be due to some confounding extraneous or “third” variable (the lack of
alternative explanation condition).
It is also important to
establish a theoretical explanation and then empirically test
predictions based on that theoretical explanation.
11.1. Why is experimental
research much stronger than nonexperimental research when the researcher is
interested in making cause and effect statements?
The strongest experimental
studies are the best for making statements of cause and effect (i.e.,
experiments with random assignment) because they establish a relationship,
proper time order (because the IV is manipulated and then the outcome is
observed), and they rule out alternative explanations (because random
assignment equates the groups on all extraneous variables at the start of the
experiment). On the other hand, nonexperimental research is good at
establishing that a relationship is present but has problems with the other two
conditions (establishing proper time order and ruling out alternative
explanations).
11.2. Why must a researcher
sometimes conduct nonexperimental research rather than
experimental research?
Because there are many
independent variables (needing to be studied) that cannot be manipulated or it
would be unethical to manipulate them. Remember that it is the research
question that drives research.
11.3. Why must researchers
watch out for the “post hoc fallacy”?
Because it is easy to
“explain” things after the fact. You should avoid the logical fallacy of
assuming that because event A preceded B, A must have caused B. If you do
generate theory or new hypotheses based on what you observed in the past (which
is fine), you must remember that you must also test those predictions with
new data.
11.4. Name of a potential
independent variable that cannot be manipulated.
Cigarette smoking, amount of
violence seen on television, whether someone drops out of high school, or
whether someone uses illicit drugs.
11.5. Explain the problems with
the simple cases of causal-comparative and correlational
research. Why is a
researcher not justified in making a cause and effect claim from
these two cases?
The bottom line is that both
of these simple cases of nonexperimental research are only useful for making
statements about the observed relationship between two variables.
Neither of these two simple cases is useful for establishing evidence of
causality; to do this, you must also attempt to establish proper time order and
to control for potentially confounding variables. In short, to conduct high
quality nonexperimental research, you must improve upon these “simple cases.”
·
(Note
that when you move into higher quality nonexperimental research, you should drop
the correlational and causal-comparative terminology and use the clearer and
more useful terminology explained in
this chapter. The nine major nonexperimental research designs that are
discussed in this chapter and are based on our terminology are shown in Table
11.3.)
11.6. Explain exactly how
strong experimental research fulfills each of the three
necessary conditions
for cause and effect.
In the strongest of all the
experimental designs you have manipulation, a comparison group, and random
assignment to groups. (By now, you BETTER have the three necessary conditions
for causation memorized.) Condition one (relationship condition) is established
in these strong designs (also called randomized designs) by checking to see if
the group means are different on the dependent variable after administration of
the independent variable. Condition two (proper time order condition) is
established because the researcher manipulates the independent variable and
then looks for changes in the dependent variable. Random assignment to groups
helps to clearly establish the third condition for causality (lack of
alternative explanation condition). Random assignment does this by equating the
groups on all known and unknown extraneous variables at the onset of the
experiment.
11.7. On which of the three
necessary conditions for cause and effect is nonexperimental
research especially
weak? On which one of the three
necessary conditions is
nonexperimental
research strong?
Nonexperimental research is
especially weak on condition three because it is always possible that an
observed relationship is a spurious (i.e., non-causal) relationship (i.e., a
spurious relationship is due to the operation of a third variable). Nonexperimental research is very strong on
condition one; that is, it is very good at showing that two variables are
related.
11.8. Explain why you cannot
make a defensible “causal claim” based on an observed
relationship between
two variables (e.g., gender and achievement) in
nonexperimental
research.
Just because two variables
are related does not mean that changes on one variable CAUSE changes in the
other variable.
Remember the three necessary conditions for establishing cause and effect? You
know that showing a relationship is only the first of the three necessary
conditions. Many simple relationships are actually spurious relationships; for
example, the following variables are positively but spuriously related: number
of fire trucks responding and amount of fire damage, ice cream consumption and
deaths by drowning, number of police officers and number of crimes, teachers
salaries and the price of liquor.
11.9. What is the purpose of
the techniques of control in nonexperimental research?
The overriding purpose is to
“control for” potentially confounding or third variables identified by the
researcher. These techniques are used to improve the two simple cases that we
discussed above (see question 11.5).
The aim of the control techniques is to help with “condition three” (the
lack of alternative explanation condition) of the three necessary conditions
for establishing cause and effect.
·
The
different approaches to control discussed are matching (i.e., selecting
participants or forming comparison groups so that the independent variable and
the extraneous variable are uncorrelated), holding the extraneous variable
constant (i.e., turning an extraneous variable into a constant by limiting
the study to one level of the extraneous variable), and statistical control
(i.e., using statistical procedures such as partial correlation, analysis of
covariance, and multiple regression to control for variables).
11.10. Which form of
nonexperimental research tends to be the best for inferring cause
and effect:
cross-sectional research, trend studies, panel studies (i.e., prospective
studies), or
retrospective research studies? Why?
The panel study (one of the
types of longitudinal research) tends to be the best because use of this design
helps the researcher establish proper time order which is not possible in
cross-sectional research. Panel studies are best when done in combination with
one or more of the control techniques (e.g., statistical control, matching)
which helps with the third necessary condition for cause and effect. Carefully
done causal-modeling (i.e., developing a hypothesized theoretical model and
then empirically testing it) can also provide moderately good evidence of
causality (e.g., not nearly as good as a strong experiment but much better that
the simple cases of nonexperimental research where only the relationship
between two variables is examined).
11.11. Explain the difference
between a direct effect and an indirect effect.
A direct effect is
the effect of the variable at the origin of an arrow on the variable at the
receiving end of the arrow (e.g., in the diagram X----->Y, X is hypothesized
to affect Y). An indirect effect is an effect occurring through an
intervening variable (e.g., in the diagram
X----->I----->Y, X is
presumed to affect Y indirectly through, or by way of, the intervening variable
labeled I).
11.12. List an advantage and a
disadvantage of causal modeling.
An important advantage is
that it requires the researcher to make explicit his or her theoretical model
(which usually includes direct and indirect relationships) and then to collect
data to empirically test the model. A disadvantage is that theoretical models
of this sort are usually tested with nonexperimental data (i.e., in studies
without manipulation or random assignment), which makes it difficult to
establish conditions two and three or the three necessary conditions.
Here is one more study
question that I think is real important:
11.13. What are the two dimensions used to classify
nonexperimental research into
nine designs?
Here are the two dimensions:
1.
Research
objective: descriptive, predictive, or explanatory.
2.
Time
dimension: retrospective, cross-sectional, and longitudinal.
These two dimensions are
crossed to form a matrix showing nine nonexperimental research designs in Table
11.3:
