Transcript Research Methods - Washington State University

Research Methods
Experiment
Components of Experiments
• Independent and Dependent variables
• Pre-testing and post-testing
• Experimental and control groups
Babbie 2005
Independent and dependent variables
• Treatment or Independent Variable the
stimulus, manipulation, or intervention that the
researcher creates or delivers to one set of
participants or clients.
• Dependent Variable is the outcome or
condition that may change as a result of being
subjected to or exposed to the treatment or an
independent variable.
Pre-testing and post-testing
• Pre-test is the measurement of the
dependent variable prior to treatment or
intervention.
• Post-test is measurement of the
dependent variable after the treatment/
independent variable has been applied.
Experimental and Control groups
• Must be as similar as possible
• An Experimental Group are those who
receive the treatment or are exposed to the
independent variable under study.
• The Control Group are those who do not
receive the treatment or independent variable
under study.
Copyright @ Allyn & Bacon 2003
Classical Experiment
• At least two groups (control and experimental)
• Randomly assign people to groups
• Treat the experimental group by manipulation
the independent variable
• Observe the effect of the treatment on the
dependent variable in the experimental group
• Compare the dependent variable differences in
the experimental and control groups
Diagram of Classical
Experimental Design
Babbie 2005
Some Notation
• Independent variable/Treatment is X
• Measurement of the dependent variable is notated
as O
• Pre-test is
O
(Observation on the dependent
variable at time 1)
• Post-test
1
O
2
(Observation on the dependent variable
at time 2)
O
O
1
1
X O
O
2
2
Relationships Among Pre-Post
Designs
• There are three major types of pre-post
program-comparison group designs all
sharing the basic design structure shown in
the notation above:
• The Randomized Experimental (RE) Design
• The Nonequivalent Group (NEGD) Design
• The Regression-Discontinuity (RD) Design
• The designs differ in the method by which
participants are assigned to the two groups.
The Randomized Experimental (RE)
Design
• In the RE, participants are assigned
randomly
O1 X O 2
O1
O2
The Nonequivalent Group (NEGD)
Design
• In the NEGD, assignment of participants
is not explicitly controlled -- they may
self select into either group, or other
unknown or unspecified factors may
determine assignment.
The Regression-Discontinuity (RD)
Design
• In the RD design, they are assigned
using a cutoff score on the pretest.
The Solomon Four-Group Design
• Two of the groups receive the treatment and two do
not.
• Further, two of the groups receive a pretest and two
do not
O
O
1
1
X O
O
X O
O
2
2
2
2
The Minneapolis Domestic
Violence Experiment (1983)
• Goal was to find the most
effective strategy
• Three groups: two treatment groups and one
control
• Police officers volunteering to take whatever
action was dictated by a random system:
instruction in an envelope
• Three different instructions: (1) arrest the
suspect; (2) separate or remove the suspect
from the scene for 8 hours; (3) advise and
mediate
Minneapolis Domestic
Violence Experiment
Experimental
group I
Experimental
group II
Control group
Arrest
Separate
Mediate
O1
X1
O2
19%
O1
X2
O2
33%
O2
37%
O1
1. Victims have interviewed every two weeks for the next 6
months, police records have been monitored as well
2. Most influential policy experiment
3. Arrest works more effectively in deterring domestic
violence
Experiments in Social Science
• Not always possible (ethical issues)
• Quasi-experiments or natural experiments
• Example: Effect of the decision to conduct
crackdown on drinking and driving by a local
police force (planned interventions)
• Occasionally, natural events (catastrophe or
tornadoes) might substitute planned
interventions
Field Experiments
• Field experiments are used outside the
laboratory.
• But although field experiments overcome the
problem of experiments taking place in an
unnatural setting, these experiments do have
other problems associated with them
• It is not possible to control variables as
closely as it is in the laboratory
Field Experiments
• Also, in some field experiments, the fact
that an experiment is taking place can
affect the results, as explained before
with the ‘Hawthorne Effect’
• To possibly avoid this, it is necessary
that the subjects of experimental
research are unaware that the
experiment is taking place
Field experiment
• This, however, raises a further problem: the
morality of conducting experiments on people
without their consent
• Some sociologists strongly object to do this
• Although field experiments open up greater
possibilities than laboratory experiments, they
are still likely to be confined to small-scale
studies over short periods of time.
• Experimentation on society as a whole, or on
large groups of society, is only likely to be
possible with the consent of governments
Field experiment
• Few governments are willing to surrender
their authority to social researchers who are
keen to test the theories and hypotheses they
have developed
• In these circumstances sociologists normally
rely upon studying society as it is, rather than
trying to manipulate it so that their theories
can be tested directly
Strengths
•
•
Weaknesses
The only method that • In real life, only rarely
one variable actually a
allows us to test the
cause of another one
causal relationships
between variables
• Difficult to test very
Random assignment of complex hypotheses
(difficult to manipulate
subjects to
and control more than
experimental and
one or two variables)
control groups allows
us to test our
• Ethical issues
hypotheses