Preview Extract
Chapter 4: Causal Designs and Marketing Experiments
True/False
1) Deterministic causation is the scientific notion that the effect always follows the cause.
Answer: False
The notion of deterministic causation is that the effect follows the cause; probabilistic
causation is the scientific notion that the effect is only probable.
2) The common sense notion talks of proving that X causes Y; the scientific notion holds that
we can only infer causality without proving it.
Answer: True
From a scientific standpoint, causality can be inferred but never really proven.
3) Due to relationships among variables caused by other factors, such as reverse causation,
omitted variables, and insufficient variation, concomitant variation is not sufficient to make
valid causal inferences.
Answer: True
Just because variables display a high level of correlation does not mean one caused the other.
4) For causation to occur, the cause must occur prior to the effect.
Answer: False
The cause can also occur simultaneously with the effect, as well as before the effect.
5) The measures taken on the test units are the independent variables in an experiment.
Answer: False
The measures are the dependent variables; the treatments are the independent variables.
6) Randomizing treatments to test units is not an effective method of handling extraneous
variables, because it does not in any way control extraneous variables.
Answer: False
Randomization is important for controlling extraneous variables in experimental design.
7) If physical control, randomization, and research design features do not eliminate the
differential effects of extraneous variables among treatment groups, the experiment is said to
be confounded.
Answer: True
When the effects of extraneous variables cannot be eliminated, then the experiment is
confounded and no causation can be assumed.
8) An experimental design involves the specification of treatments to be manipulated, test
units to be used, and independent variables to be measured.
Answer: False
An experimental design must specify treatments to be manipulated (i.e., the independent
variables), test units to be used, the dependent variables to be measured, and procedures for
dealing with extraneous variables.
9) External validity is concerned with whether the experimental results can be generalized.
Answer: True
External validity is concerned with whether the results can be generalized.
10) Extraneous variables need to be controlled to ensure that the experiment is externally
valid.
Answer: False
Controlling extraneous variables will assure researchers of internal validity, not external
validity.
11) The R in X-O-R syntax stands for “reference,” which are the test units being observed.
Answer: False
R indicates that individuals have been assigned at random to separate treatment group.
12) In experiments, the greater the time between observations (O1 and O2), the greater the
chance of a history or maturation effect confounding the experiment.
Answer: True
Time between observations increases the possibility of both history effect and maturation
effect.
13) The reactive (or interactive) testing effect occurs when the first observation affects the
second observation.
Answer: False
This is direct (or main) testing effect, not reactive or interactive testing effects.
14) Test unit mortality refers to instances where researchers exceed the treatment and
observation deadlines specified in the experimental design.
Answer: False
Test unit mortality refers to test units withdrawing from the experiment while it is in progress.
15) The one-group pre-test-post-test design is represented symbolically by “O1 X O2.”
Answer: True
This pre-experimental design uses a pre-test before the treatment and a post-test after the
treatment.
16) A True experimental design is one where the researcher is able to eliminate all extraneous
variables as competitive hypotheses to the treatment, at least in theory.
Answer: True
True experimental designs, at least in theory, rule out all of the threats to internal validity.
17) In True experimental designs, extraneous variables must never be present in the treatment
group(s), although they can be present in control group(s) if properly controlled.
Answer: False
Extraneous variables can be present as long as they operate equally on all treatment and
control groups. Confounding results occur when extraneous variables operate differently
among treatment and control groups.
18) In a Solomon four-group experimental design, all extraneous variables and the interactive
testing effect can be controlled, but their effects cannot be measured.
Answer: False
Not only can extraneous variables and the interactive testing effect be controlled, they can
also be isolated and measured.
19) A time-series experiment undertakes periodic measurement on dependent variables for
test units. The treatment is then introduced, or occurs naturally, and the periodic
measurements continue on the same test units to monitor the effects of the treatment.
Answer: True
This is the definition of a time-series experiment.
20) The primary difference between the time-series experimental design and the multiple
time-series design is the randomization of the test units to the treatment group in the multiple
time-series design.
Answer: False
The primary difference is the presence of a control group with the multiple time-series
design.
21) Only experimental and quasi-experimental designs can identify causal relationships.
Answer: True
Unfortunately, descriptive studies are often used to argue causal relationships, but only
experimental or quasi-experimental designs can rigorously and reliably separate cause and
effect.
22) Laboratory settings generally offer lower control over confounding variables than other
studies, but the results are more able to be generalized.
Answer: False
It is the reverse: laboratory settings generally offer greater control over confounding variables
than other studies, but the results are less able to be generalized.
23) In a completely randomized design, treatments are randomly assigned to groups of test
units based on an external criterion variable.
Answer: False
In a completely randomized design, treatments are assigned to test units in an entirely random
manner. In a randomized block design, test units are combined into blocks based on an
external criterion variable and then treatments are randomly assigned to those groups. By
definition, random assignment can not be based on an external criterion variable.
24) In a randomized block design, the dependent variable must be of at least ordinal scale.
Answer: False
In a randomized block design, blocks are nominally scaled.
25) In the Latin square (LS) design, each row and column category defines a block situation,
which means that each treatment must appear once in each row and once in each column.
Answer: True
Although treatments must be assigned randomly, they must satisfy this restriction of only one
per row and one per column.
MULTIPLE CHOICE
1) Under which conditions can researchers claim a causal inference?
a. concomitant variation
b. high level of correlation between variables
c. time order of occurrence of variables
d. elimination of other possible causal factors
e. a, b, and c only
f. a, c, and d only
Answer: False
Causal inferences require establishing concomitant variation, time order of occurrence of
variables, and elimination of other causal factors. A high level of correlation between
variables does not necessarily imply causation.
2) The extent to which a cause, X, and an effect, Y, occur together or vary together in the way
predicted by a hypothesis under consideration is
a. concomitant variation
b. convergent variation
c. deterministic causation
d. interaction effect
e. time occurrence of variables
Answer: A
By definition, concomitant variation is the extent two variables occur together or vary
together as hypothesized.
3) In addition to causation, the other types of relationships possible among correlated
variables include all of the following except
a. reverse causation
b. omitted variables
c. sequential effect
d. insufficient variation
Answer: C
In addition to causation, other types of relationships that could explain correlation between
variables are reverse causation, omitted variables, and insufficient variation.
4) One of the most common and pernicious erroneous assumptions about causation in
marketing research is
a. not considering reverse causation
b. not eliminating other variables as the cause
c. not establishing sufficient variation
d. not ruling out the sequential effect
Answer: B
One of the most common and pernicious errors in marketing research is the failure to test for
and systematically exclude other possible variables as the cause of concomitant variation.
5) In True and quasi-experiments, treatments need to form
a. a ratio scale
b. at least an interval scale
c. at least an ordinal scale
d. any of the four types of scales
Answer: D
Treatments can be measured using any of the four types of scales.
6) The _______________ are the entities to whom (or to which) the treatments are presented
and whose response to the treatments is measured.
a. population units
b. researchers
c. test units
d. confounds
Answer: C
Test units are the entities which are presented and whose response to treatments is measured.
7) The _______________ variables are all variables other than the treatments that potentially
affect the response of the test units to treatments.
a. dependent
b. extraneous
c. independent
d. control
Answer: B
Extraneous variables can distort the dependent variable measures in such a way as to weaken
or entirely invalidate a researcher’s ability to make causal inferences.
8) All of the following are methods for controlling the effects of extraneous variables in
experiments except
a. if practical, hold such extraneous variables constant
b. assign treatments to test units randomly
c. use experimental designs that control for the effects of extraneous variables
d. eliminate the effect of extraneous variables on test units
Answer: D
The effect of the variables on test units remains; potential extraneous variables must be
controlled for in one of the other three ways listed.
9) If physical control, randomization, and design features do not eliminate the differential
effects of extraneous variables among treatment groups, the experiment
a. has been confounded, but causal statements are still possible
b. has been confounded, and no causal statements are possible
c. must be repeated without the presence of extraneous variables
d. loses external validity, but can still salvage internal validity
Answer: B
If the effects of extraneous variables cannot be eliminated, then the experiment has been
confounded and the researcher cannot imply any causation.
10) There is one line of defense against a confounding variable; its effects on the dependent
variable may be statistically controlled through a technique called
a. analysis of variance
b. analysis of covariance
c. regression analysis
d. static-group comparison design
Answer: B
Analysis of covariance (ANCOVA) can control for a confounding variable if the researcher is
aware of it and can measure it.
11) _______________ is the basic minimum _______________ that must be present in an
experiment before any conclusion about treatment effects can be made.
a. External validity, validity
b. Internal validity, validity
c. External reliability, reliability
d. Internal reliability, reliability
Answer: B
Without internal validity, the experiment is confounded and largely useless for the purpose of
causal inference.
12) The question of whether the observed effects on the test units could have been caused by
variables other than the treatment refers to:
a. external validity
b. internal validity
c. possible instrumentation effects
d. possible concomitant variation
Answer: B
Internal validity relates to whether the observed effects on the test units could have been
caused by variables other than the treatment, that is, by extraneous variables.
13) In experimental design syntax, _______________ represents the exposure of a test group
to an experimental treatment.
a. E
b. O
c. R
d. X
e. the subscript of O
Answer: D
“X” represents the exposure to an experimental treatment; “O” refers to an observation; and
“R” indicates random assignment to a treatment.
14) The _______________ effect refers to the occurrence of specific events that are external
to the experiment but that take place at the same time as the experiment.
a. history
b. maturation
c. testing
d. instrumentation
e. statistical regression
Answer: A
History effects refer to events that are external to the experiment but can affect the dependent
variable.
15) Getting older, growing hungrier, and developing fatigue are examples of
_______________ effects, which is changes in the experimental units themselves that occur
with the passage of time.
a. history
b. maturation
c. testing
d. instrumentation
e. statistical regression
Answer: B
Maturation effects are similar to history effects except that they are concerned with changes
in the experimental units themselves that occur with the passage of time.
16) The _______________ effect is concerned with the possible effects on the experiment of
taking a measure on the dependent variable before presentation of the treatment.
a. history
b. maturation
c. testing
d. instrumentation
e. statistical regression
Answer: C
The testing effect occurs when measurement of the dependent variable occurs before the
treatment.
17) _______________ refers to the assigning of test units to treatment groups in such as way
that the groups differ on the dependent variable before the presentation of the treatments.
a. The main testing effect
b. Test unit mortality
c. Selection bias
d. The instrumentation effect
Answer: C
If the test units already differ in terms of the dependent variable before the treatment is given,
then selection bias has occurred.
18) A single group of test units is first exposed to a treatment, and then a measurement is
taken on the dependent variables. Which experimental design is this?
a. one-shot case study design
b. one-group pre-test-post-test design
c. static-group comparison design
d. pre-test-post-test control group design
Answer: A
The one-shot case study design consists of a treatment and an observation.
19) All of the following are pre-experimental designs except
a. one-shot case study design
b. one-group pre-test-post-test design
c. static-group comparison design
d. pre-test-post-test control group design
Answer: D
The pre-test-post-test control group design is a True experimental design, not a preexperimental design.
20) The _______________ pre-experimental design uses two treatment groups, one that has
been exposed to the treatment and one that has not.
a. one-shot case study
b. one-group pre-test-post-test
c. static-group comparison
d. pre-test-post-test control group
Answer: C
The static-group comparison uses two groups, one of which is exposed to the treatment. Both
groups are then observed at the same time after the one has been exposed to the treatment.
21) The _______________ design controls for all extraneous variable effects on internal
validity, plus the interactive testing effect.
a. Solomon four-group
b. one-group pre-test-post-test
c. Latin square
d. pre-test-post-test control group
Answer: A
In the Solomon four-group design, an additional control group and experimental group are
added that do not receive the pre-test, thus effectively controlling for the interactive testing
effect.
22) The interactive testing effect cannot occur with the post-test-only control group
experimental design, because
a. both the control and treatment group are measured after the treatment
b. there is no pre-test in the design
c. the testing units are randomly assigned to the treatment groups
d. the control group and treatment group do not interact
Answer: B
Because there are no pre-tests, there cannot be any interactive testing effect.
23) In a quasi-experimental design, researchers normally lack control of all of the following
except
a. when the treatments are scheduled
b. which test units receive the treatment
c. data collection and/or measurement procedures
d. ability to randomize the test unit’s exposure to treatments
Answer: C
With quasi-experimental design, researchers only have control over data collection
procedures.
24) The fundamental weakness of the time-series quasi-experimental design is the
researcher’s inability to control for the _______________ effect.
a. statistical regression
b. history
c. main testing
d. instrumentation
Answer: B
Because of repeated measures over time, the history effect can confound the results.
25) The biggest problem with the equivalent time-sample quasi-experimental design is the
possibility of the _______________ effect.
a. interactive testing
b. history
c. statistical regression
d. instrumentation
Answer: A
Because treatments and measurements are repeated, there can be some learning from one
period to another.
26) In the _______________ quasi-experimental design, both the experimental group and the
control group are given pre-test and post-test measurement, but the two groups do not have
pre-experimental test unit selection equivalence, i.e. the control group cannot be considered
equivalent to the test group.
a. time-series
b. multiple time-series
c. nonequivalent control group
d. equivalent time-sample
Answer: C
This is the definition of the nonequivalent control group design.
27) Only _______________ can identify causal relationships.
a. descriptive studies
b. field experiments
c. laboratory experiments
d. experimental and quasi-experimental designs
Answer: D
Both experimental and quasi-experimental designs can identify causal relationships.
28) Which of the following statements comparing laboratory environments to field studies is
false?
a. Laboratory environments provide the researcher with maximum control over possible
confounding variables.
b. Laboratory environments are generally less expensive than field studies.
c. The complex nature of laboratory experiments means that they require more time to
conduct than field studies.
d. The laboratory environment provides greater internal validity than do field studies.
Answer: C
Laboratory experiments tend to be simpler in nature and, therefore, are usually less time
consuming to execute. They do offer maximum control over extraneous variables, thereby
providing greater internal validity, and they are generally less expensive than field studies.
29) When comparing laboratory experiments with field experiments, field experiments offer
an advantage over laboratory experiments in terms of
a. internal validity
b. external validity
c. costs
d. speed
Answer: B
Laboratory experiments excel in internal validity, cost and speed, whereas field experiments
offer greater external validity.
30) All of the following are limitations of experimentation except
a. it is not always possible to control the effects of extraneous variables
b. experiments cannot control for all sources of internal invalidity
c. care must be taken that the experimenter does not say and do anything that might bias
responses
d. in field experiments, lack of cooperation can limit experimental activity
Answer: B
True experimental designs can control for all sources of internal invalidity, and most of the
quasi-experimental designs can as well.
31) In controlling internal sources of invalidity, True experimental designs control for
a. all of the sources of internal invalidity
b. all of the sources of internal invalidity except selection bias and test unit mortality
c. selection bias, instrumentation, and interactive testing effects only
d. selection bias and test unit mortality only
Answer: A
One of the primary advantages of True experimental designs is they control for all of the
internal sources of invalidity.
32) Exposing two random samples of visitors to a web page to alternate versions of a web
page in order to optimize website elements is an example of
a. a factorial design
b. a nonequivalent control group design
c. A/B testing
d. a completely randomized design
Answer: C
A/B testing involves an experiment where traffic to a website is randomly redirected to two
different versions of the website simultaneously, the current version and an alternate version.
This method is widely used to optimize a variety of website elements.
33) The simplest type of designed experiment is the
a. completely randomized design
b. randomized block design
c. Latin square design
d. factorial design
Answer: A
The simplest is the completely randomized design because it involves investigating only one
independent variable.
34) The _______________ is useful when the researcher is investigating the effect of one
independent variable.
a. completely randomized design
b. randomized block design
c. Latin square design
d. factorial design
Answer: A
Wit the completely randomized design, only one independent variable is investigated.
35) The _______________ is built on the principle of combining test units into blocks based
on an external criteria variable that might impact the results of the study.
a. completely randomized design
b. randomized block design
c. Latin square design
d. factorial design
Answer: B
If one or more extraneous variables is assumed not to be constant, then a randomized block
design (RBD) can be used to create blocks based on the extraneous variable of concern.
36) In the randomized block design (RBD), researchers
a. can use only one blocking factor
b. can use one or two blocking factors
c. can use multiple blocking factors as long as it matches the number of dependent variables
d. cannot use any blocking factors
Answer: A
With the RBD design, only one blocking variable can be used. If there are more external
variables, the blocking factor must be defined in categories involving more than one variable.
37) When researchers wish to control for the effects of two extraneous variables, a
_______________ may be used.
a. completely randomized design
b. randomized block design
c. Latin square design
d. either b or c
Answer: C
The Latin square design allows for two extraneous variables.
38) In a Latin square (LS) design the number of categories of each extraneous variable to be
controlled
a. must not be greater than two
b. exactly equals the number of treatments
c. has to be less than the number of treatments minus one, divided by two [<(n – 1)/2]
d. exactly equals the number of dependent variables
Answer: B
The number of treatments must be equal to the total number of categories of the extraneous
variables.
39) If researchers want to examine two or more independent variables in an experimental
design, they must use a
a. pre-test-post-test control group design
b. randomized block design
c. Latin square design
d. factorial design
Answer: D
Only the factorial design allows for more than one independent variable.
40) To measure the interaction of independent variables, researchers would use a
a. pre-test-post-test control group design with the interaction observed before and after
treatment
b. randomized block design with the interaction treated as the blocking factor
c. Latin square design with the interacting independent variables treated as levels of
extraneous variables
d. factorial design
Answer: D
Only the factorial design allows breaking effects down to individual effects, main effects, and
interaction effects of more than one independent variable. The RBD presumes that there is no
interaction between the blocking factor and the independent variable, and the LS design
assumes that there is no interaction between the two blocking factors. Only the factorial
design allows measurement of the effect of interaction between independent variables.
SHORT ANSWER
1) Explain what is meant by the term “experiment.”
Answer: An experiment is carried out when one or more independent variables is deliberately
manipulated or controlled by the experimenter in a planned fashion and the effects on the
dependent variable are measured while controlling for other variables that might confuse the
ability to make causal inferences.
2) What elements are in an experimental design?
Answer: An experimental design involves the specification of treatments to be manipulated,
test units to be used, dependent variables to be measured, and procedures for dealing with
extraneous variables.
3) Compare and contrast laboratory experiments and field experiments in terms of validity,
costs, and time.
Answer: Laboratory experiments have a higher level of internal validity, tend to cost less, and
normally take less time. Field experiments offer a higher level of external validity.
4) Identify and define the symbols used in experimental syntax.
Answer: “X” represents the exposure of a test group to an experimental treatment.
“O” refers to processes of observation or measurement of the dependent variable on the test
units.
“R” indicates that individuals have been assigned at random to separate treatment groups.
5) Define the reactive testing effect.
Answer: The reactive testing effect occurs when the test unit’s pre-treatment measurement
affects the reaction to the treatment. (Also called the interactive testing effect)
ESSAY
1) Causation is not the only reason variables can be correlated. Discuss the other possible
explanations and give an example of a causal statement along with examples of the other
possible explanations.
Answer: Other reasons for variables to be correlated include reverse causation, omitted
variables, and insufficient variation. Student examples will vary. One example of a causal
statement is “Positive attitudes toward the brand caused an increase in sales.” Reverse
causation for this would be the increase in sales causing positive attitudes toward the brand;
omitted variables could be that the product design was improved during the period of
measurement; insufficient variation would be that attitudes improved, but not enough to
account for the sales increase.
2) Discuss the courses of action a researcher can follow to control extraneous variables and
the statistical recourse that may be available if these fail to prevent a confound.
Answer: The three courses of action to control extraneous variables include 1) holding
extraneous variables constant where practical, 2) assigning treatments to test units randomly,
and 3) using experimental designs that control for the effects of extraneous variables. If an
extraneous variable confounds the experiment, analysis of covariance (ANCOVA) can be
used to statistically de-bias the dependent variable by accounting for it separately within each
of the levels of the confounding variable.
3) Identify the three pre-experimental designs and the syntax that is used to represent them.
Describe how each is conducted.
Answer: 1) one-shot case study design:
A single group of test units is first exposed to a treatment X, and then a measurement is taken
on the dependent variables.
2) one-group pretest-posttest design:
A pretest is taken before the experimental treatment and then compared to the posttest results
after the treatment.
3) static-group comparison design:
Two treatment groups are used, one that has been exposed to the treatment and one that has
not.
4) Given an experiment with the following syntax:
Discuss what kind of design this syntax represents; whether it is pre-experimental, quasiexperimental, or True experimental; and which extraneous variables it controls for. Explain
how treatment effects can be separated from other effects.
Answer: This design is called a Solomon four-group design. It is a True experimental design
that controls for all extraneous variables. Where TE is the treatment effects, EXT is the
extraneous effects, and IT is the interactive testing effect,
O2 – O1 = TE + EXT + IT
O4 – O3 = EXT
O5 – O1 = TE + EXT
O5 – O3 = TE + EXT
O6 – O1 = EXT
O6 – O3 = EXT
This allows solving for treatment effects separately from extraneous effects and the
interactive testing effect. Simply,
TE = O5 – O6
IT = O2 – O5
EXT = O6 – O1 or O6 – O3 or O4 – O3
However, to maximize statistical power and minimize the potential for bias, the results of the
four groups are generally used equally.
5) Discuss the steps in properly conducting an experiment.
Answer: The following steps should be followed:
1) State the problem clearly
2) Formulate a hypothesis
3) Construct an experimental design
4) Formulate mocked-up data and use them to check to see that they meet the specifications
in the problem statement
5) Check that the types of experimental results possible can be analyzed by available
statistical procedures
6) Perform the experiment
7) Apply statistical analysis procedures to the results to determine whether effects are real
8) Draw conclusions, paying attention to both internal and external validity
Test Bank for Modern Marketing Research: Concepts, Methods, and Cases
Fred M. Feinberg, Thomas Kinnear, James R. Taylor
9781133188964, 9781133191025, 9780759391710
