Unit 10: Quasi Experimentation Flashcards
quasi experiments
The researcher selects participants from pre-existing groups (does not have control over assigning subjects to conditions).
Sometimes called ex post facto because assignment happens after the groups have formed (ex. male/female)
Difference between groups may not be the cause of the effect, could be obscuring the effect.
non-equivalent control group designs
Experiment with a control and experimental group, but P are nor randomly assigned to them.
repeated treatment designs
A design that attempts to improve validity by introducing the treatment more than once (return to baseline in between).
interrupted time-series designs
A design that considers the trend before and after the experiment, so one group can be tested alone (no control)
program evaluation
A set of techniques for determining the effectiveness of a social service program.
summative evaluation
A way to determine whether a program is functioning adequately, meeting its goals.
Often after completion.
Can be used to determine cutbacks, changes, expanding, etc.
formative evaluation
The goal is to improve the program.
cross-sectional study
A design that tests different age groups at the same time to determine age-based differences.
longitudinal study
A design that follows one P/group and tests them at different points over time to determine age-based differences.
cross-sequential design
Tests 2+ groups at 2+ times to control for cohort effects, secular effects.
Cohort effect
Effects of characteristics that are specific to a cohort.
secular effect
Effects due to changes in the larger population/social environment.
How are quasi experiments different from true experiments?
Still intend to show cause-effect, but one of these is not met:
- researcher randomly assigns P to 2 groups that are roughly equivalent
- researcher manipulates the IV
- there is a control or baseline condition to compare to experimental/treatment condition
Under what circumstances can quasi experiments be a useful tool?
- when only option/better than nothing
- when can’t practically or ethically assign P to groups for exposure to the IV
- can test lab results for external validity
- can suggest further research for lab
- can have higher external validity than true if more closely resembles the real world
What is the major weakness of quasi experiments?
Confounding variables. Because the researcher doesn’t have complete control over assigning P to conditions. Difference between groups could be the cause of effect, or could be obscuring the real cause.
Major threats in field testing (need to the diligent about these in quasi-experiments)
- diffusion/imitation of treatment - groups share info/resources with each other - reduces difference
- compensatory equalization - workers give treatment to control P - reduces differences
- compensatory rivalry - disadvantaged group is motivated to perform better - reduces difference
- resentful demoralization - disadvantaged group is discouraged and performs worse - increases difference
What is a non-equivalent control group design with pretest and post-test? Under what circumstances are the results of quasi-experiments that use this design interpretable?
Control and experimental groups are not roughly equal - they differ in some way.
Add in a pre-test and post-test and then compare results.
The results are interpretable if they can be attributed to the difference between groups.
Pre-test scores start similar, then experimental change but control not, OR experimental starts lower and ends up higher.
Under what circumstances is an interrupted time-series design useful?
When no control group is possible.
When data for before and after can be accessed to show trends.
What different strategies are used in non-equivalent control group designs and interrupted time-series designs to control for confounding variables?
- comparing pre- and post-test results with other groups
2. considering trends before and after the test
What are repeated-treatment designs? How do these designs control for confounding variables?
The treatment is introduced multiple times, with a return to baseline between each. If the effect is consistent each time, it increases confidence that the IV is causing the effect in the DV. Proves the effect isn’t due to a trend.
Which quasi-experimental designs can be used with a single subject (an individual, a company, or some other units)?
- interrupted time series
- repeated treatments
- longitudinal
What are the common quasi-experimental designs used in developmental psychology?
- longitudinal
- cross-sectional
- cross-sequential
advantages and disadvantages of Cross-sectional study
A: faster/cheaper, done by one researcher
D: hard to control for cohort effects, age is confounded with DOB.
advantages and disadvantages of Longitudinal study
A: avoids cohort effects
D: takes longer, more expensive, more than one researcher, confounds age with time of testing (secular effects)
advantages and disadvantages of Cross-sequential design
A: addresses the confounding in cross-sectional and longitudinal. Shows secular effects, though does not control for them.
How can quasi-experimental designs be used in evaluations of social programmes?
They are useful because P are often assigned (not randomly) to groups. Comparing existing groups according to an IV requires a quasi-experimental design.
