Transcript Research Design - Blogs @ Suffolk University

Research Design:
Alan Monroe: Chapter 3
The Concept of Causality (31)
The types of research designs reviewed here are all intended
to test whether one variable causes another or causes
variation in another.
Three (3) Requirements of Causality (31-32)
Correlation: two things tend to occur at the same time (not
sufficient to est. causation)
Examples:
Whenever there is a foreign policy crisis, presidential
popularity increases.
If Catholic, then more likely to oppose abortion.
Time Order: cause has to happen before the effect.
Non-Spuriousness: to make sure any correlation we observe
between the independent and dependent variables is not caused
by other factors.
Types of Research Designs (32)
1. True Experimental Design
It involves a group of subjects (units of analysis), which is
divided into two groups (randomly, to assure they are
identical on the DV).
Experimental and Control Groups
The first group is the experimental group, the second is the
control group. The experimental group receives a stimulus
(the Independent Variable), the control does not.
Post-Test
A Post-Test is then given to both groups to test the effect (DV)
of the stimulus (IV). You then compare the results.
…
True Experimental Design
Examples of Experimental Design (33)
Introduction to American Government Example: Does it
Increase Political Interest? (see Chart on p. 33)
Hypothesis: taking course increases political interest in
college students.
…
Problems With True Experimental Design
Hard to get truly Representative Samples (hard to get accurate
sample of an entire population, one solution is to reduce size of
population: college students for example.)
Artificial setting (does it test real behavior)
Outside Influences (you can never fully isolate subjects from
other variables.)
Ethical Considerations (cannot mistreat or expose humans to
harmful stimuli)
…
The Quasi Experimental (Natural Experiment) (37)
2) Quasi Experimental It is also called the before and after test:
you compare the DV (a Pretest and Posttest) before and
after the IV has been applied.
Differs from Experimental Design in several ways:
Groups are not assigned (we observe some happen, and
then go back and sort into experimental and control
groups.)
Requires a Pretest of DV so amount of change can be measured.
…
Quasi Experimental Design
Quasi Experimental:
Presidential Debate Example (38)
Hypothesis: watching a presidential debate increases intensity
of support for the candidate.
Subjects: students in a class
Pretest: before debate give them a survey measuring their
attitudes about the candidates
Posttest: did they watch the debate, and what is the strength of
their preference.
Meeting Conditions of Causality:
Quasi Experimental (38)
Correlation: change between pretest and post-test has to be
significant (indicating IV had an effect)
Time Order: includes measure of DV before and after IV.
Non-Spurious: effect of all outside forces is theoretically equal
on all subjects. (they are all exposed to same amount of TV
ads, thus any changes comes from the IV).
…
Correlational Design (40)
It is very simple: collecting data on the IV and DV in order to
see if there is a pattern or relationship. It is the most common
design in political science.
Examples:
Turnout in Urban Areas
IV: urbanization
DV: voter turnout
Operational Definitions:
Urbanization: percentage of pop. Living in “urban places,”
according to US Census.
Turnout: votes cast divided by voting-age population.
Correlational Design
Meeting Conditions of Causality:
Correlational Design (38)
Correlation: is directly tested between the IV and DV.
Time Order: it is weakest here: there is no consideration for the
point in time when the IV and DV occurred. Have to reliable
on IV that are known to exist before DV, like race, gender.
Non-Spurious: considers control variables.
…