Chapter 8 Quasiexperimental and Nonexperimental Designs
- Randomization is what makes experiments true.
- In many instances, we cannot directly manipulate an independent variable such as SES, race, gender, age or
- For an experimenter either will have to select participants who have been exposed to specified events such as
war or trauma or have particular characteristics.
- Subject variable – When particular individual characteristics are used for the base of selecting research
participant (in Quasiexperiment) that can be measured or described but not directly manipulated like weight,
height or gender.
- Natural treatment – A naturally occurring event is treated as a quasi-independent variable, which can be
measured but not manipulated. Settings in the real world.
- Quasiexperiment – is one that investigates the effects of a quasi-independent variable on a dependent variable.
There is a comparison group that is comparable with the experimental group in critical ways, but participants are
not randomly assigned to the experimental and comparison groups. The word quasi means “if”. It does resemble
a true experiment.
- It is used to study things such as child abuse or gender differences.
- Two types of Quasiexperimental designs:
o Nonequivalent-control-group designs: Have experimental and comparison groups that are designated
before the treatment occurs and are not created by random assignment
o Before-and-after designs – Have a pretest and posttest but no comparison group. The participants
exposed to the treatment serve, at an earlier time, as their own controls.
Natural Treatments as Quasi-Independent Variables
- If we wanted to study psychological effects of 9/11 we would select as our quasi-independent variable 9/11.
- Ecological validity – The extent to which a study or experiment approximates the actual real life phenomenon
- Ex post facto – “After-the-fact” analysis of its effect on a particular dependent variable such as memory, suicide
rates, or grades.
Subject Variables as Quasi-Independent Variables
- If we wanted to study 9/11 but this time we are interested in gender differences in response to 9/11. Now we
have two quasi-independent variables, one we consider as natural treatment, 9/11 and the other a subject
- Mixed-factorial design – It combines both between- and within-subject factors.
- If groups formed by random assignment subsequently differed on the dependent variable, then such difference
would be attributable to treatment or independent variable.
- In quasi experiments it is not random assignment and therefore the control group can never be considered to be
equivalent to the experimental group.
- The confounds clearly limit the extent to which we can draw internally valid conclusions. - We can see that Quasiexperimental studies with non-equivalent control or comparison groups often receive
high marks for their external validity in general and their ecological validity in particular. That means they are
often high in generalizability and high in realism.
- Two methods of selection of a control group can be used to make it as comparable as possible to the experiment
o Individual matching – Individual cases in the treatment group are match with similar individuals in the
control group. Though in many studies it may not be possible.
o Aggregate Matching – It means finding a comparison group that has similar distributions on key
variables: the same average age, the same percentage female, and so on. Individuals must not be able to
choose whether to be in the treatment group or control group.
- Often a researcher can collect pretest measures and then match participants either individually or in the
aggregated on pretest scores.
- Pretest measures can become problematic for interpreting posttest scores
- Regression Artifact – It can occur anytime a pretest measure is used for matching. A threat to internal validity
that occurs when subjects who are chosen for a study because of their extreme scores on the dependent variable
become less extreme on the posttest due to natural cyclical or episodic change in the variable.
- Regression to the mean – Trend for extreme scores on a measure to move closer to the group average when
retested due to inherent unreliability of measurement. Example: If we want to calculate the mean pretest
reading comprehension score for two student groups. There are 20 students in each class, so we would sum all of
their pretest scores, divided by 20, to get the mean or average for each group. We then compared students with
extreme pretest scores (very high or very low) with posttest scores. What often occurs is that people who
received extreme scores on a test of ability when retested on that same ability earn scores that are closer to the
mean of their group than were their original scores.
- Regression to the mean may be a confounding factor for interpreting results of studies that match groups on
pretest measures. It is important to consider whenever random assignment is not or cannot be used and groups
are matched on a pretest measure.
- Common feature is absence of a comparison group. It is useful for studies of interventions that are experienced
by every case in some population such as total coverage programs like Social Security.
- Interrupted-time-series design – Is used in Quasiexperimental research to examine observations before and
after a naturally occurring treatment. Such as a public policy change on a particular behaviour like influence of
changing the age limit for drinking on car accidents in young drivers.
o You would look at accident rates over several years and need to know when the time series is
interrupted by natural treatment which in this example is law change.
- Multiple group before-and-after design – Several before-and-after comparisons are made involving the same
independent and dependent group but with different groups.
- Repeated-measures panel design – It includes several pretest and posttest observations of the same group.
- Time series designs- A Quasiexperimental design consisting of many pretest and posttest observations of the
same group (preferably 30 or more). They are useful for studies of the impact of new laws or social programs that
affect everyone and that are readily assessed by some ongoing measurement. Such as divorce rates after
Oklahoma City bombing. A Quasiexperiment of Memories of 9/11
- Sharot, Martorella, Delgado, and Phelps(2007) recruited participants who had been in the vicinity (Manhattan)
of the World Trade Center (WTC) on the morning of 9/11.
- Functional magnetic resonance imaging (fMRI) – Records the brain at work, in real time, while an individual
performs a task.
Experiment Stimuli and Task
- They used fMRI and participants viewed words that numerous prior studies have shown to be effective cues for
- The researcher wanted to see if these words would cue different memories depending on the experimental
conditions in which they were presented. There were two experimental conditions: September 2001 and
Summer 2001 and the same word cues were presented in both.
- Participants saw same word cues paired with either September 2001 or Summer 2001 and then rated each word-
cued memory on 6 characteristics.
- Sharot and colleagues used memory as a within-subjects variable with two levels: summer 2001 and
September 2001. As one of the dependent measures, the researchers recorded brain activity while participants
viewed cue words in both condition. The other dependent measure was ratings of the word cues on the six
characteristics that participants completed outside the fMRI.
Ex Post Facto Variables and Comparison Groups
- Sharon et al. found that not only did 9/11 memories have higher ratings on each of the attributed but individuals
differed considerably in their ratings.
- They examined the surveys the participants had completed and divided them into being closer to the WTC or