William M. Trochim
James P. Donnelly
Kanika Arora
10
Quasi-Experimental
Design
10.1 Foundations of Quasi-Experimental
Design
• “Quasi” means “sort of”
• Quasi-experiments have:
– A control group
– A treatment (or program) group
– Variables
• Quasi-experiments do not have:
– Random assignment to groups
10-2 The Nonequivalent Groups Design
• One of the most frequently used
quasi-experimental designs
– Looks just like a pretest-posttest design
– Lacks random assignment to groups
– As a result, the treatment and control groups
may be different at the study’s start
– Raises a selection threat to internal validity
10.2a Reaching Cause-and-Effect
Conclusions with the NEGD
10.2a Plot of Pretest and Posttest Means
for Possible Outcome 1
10.2a Plot of Pretest and Posttest Means
for Possible Outcome 2
10.2a Plot of Pretest and Posttest Means
for Possible Outcome 3
10.2a Plot of Pretest and Posttest Means
for Possible Outcome 4
10.2a Plot of Pretest and Posttest Means
for Possible Outcome 5
10.3 The Regression-Discontinuity Design
• A pretest- posttest program comparisongroup quasi-experimental design in which
a cutoff criterion on the preprogram
measure is the method of assignment to a
group
10.3a The Basic RD Design
• Notation
– C indicates that groups are assigned by
means of a cutoff score on the premeasure
– An O stands for the administration of a
measure to a group.
– An X depicts the implementation of a program
• Each group is described on a single line
10.3a Regression Line
• A line that describes the relationship
between two or more variables
10.3b The Role of the Comparison Group
in RD Designs
10.3c The Internal Validity of the RD
Design
• In principle, then, the RD design is as
strong in internal validity as its randomized
experimental alternatives
• In practice, however, the validity of the RD
design depends directly on how well you
can model the true pre-post relationship,
certainly a serious statistical challenge
10.3d Statistical Power and the RD Design
• To achieve the same level of statistical
accuracy, an RD design needs as much as
2.75 times the participants as a
randomized experiment
• Example: if a randomized experiment
needs 100 participants to achieve a
certain level of power, the RD design
might need as many as 275
10.3e Ethics and the RD Design
• RD designs tend to be more ethical,
because those who need a program or
treatment the most can receive it
10.4a The Proxy Pretest Design
• A post-only design in which, after the fact,
a pretest measure is constructed from
preexisting data
– Usually done to make up for the fact that the
research did not include a true pretest
10.4b The Separate Pre-Post Samples
Design
• A design in which the people who receive
the pretest are not the same as the people
who take the posttest
10.4c The Double-Pretest Design
• A design that includes two waves of
measurement prior to the program
– Addresses selection-maturation threats
10.4d The Switching-Replications Design
• A two-group design in two phases defined
by three waves of measurement
– In the repetition of the treatment, the two
groups switch roles:
• The original control group in phase 1 becomes the
treatment group in phase 2, whereas the original
treatment group acts as the control
10.4e The Nonequivalent Dependent
Variables (NEDV) Design
• At first, looks like a weak design
• But pattern matching gives researchers a
powerful tool for assessing causality
– The degree of correspondence between two
data items
10.4f The Regression Point Displacement
(RPD) Design
• A pre-post quasi-experimental research
design where the treatment is given to
only one unit in the sample, with all
remaining units acting as controls
– This design is particularly useful to study the
effects of community level interventions
10.4f The Regression Point Displacement
(RPD) Design
Analyze with ANCOVA
Discuss and Debate
• Why can quasi-experiments be more
ethical than randomized experiments?
• What are the strengths and the
weaknesses of quasi-experimental
designs?