Quasi-Experimental Designs
Quasi-experiments - research design types which approximate experiments and which are used when it is infeasible or unethical to use an experimental design. These designs take advantage of naturally-occurring phenomena to examine the relationship between dependent and independent variables. Quasi-experiments do not control for all other variables as well as experiments, but they often do so well enough to provide acceptable results.
Types of quasi-experimental designs:
1. Posttest only control group (post-facto design) - similar to experiments but do not include pre-test of dependent variable; similarity of pretest scores is assumed based on similarity between experimental and control groups.
2. Pretest-posttest comparison group - similar to experiments but do not include control group selected by researcher; takes advantage of non- experimental selection process which result in a comparison group sufficiently similar to experimental group to justify use as "control." May include repetition with switch, so that E. group becomes C. group and vice- versa. This latter design, sometimes called staging, is popular in evaluation research.
Example of staging: Elements are assigned to original E. group on first-come, first-served basis. E. group is enrolled in program designed to educate adolescents about impact of alcohol abuse. This group and C. group composed of those on waiting list are both pre- and post-tested. After original E. group has completed program, C. group becomes E. group through enrollment; original E. group becomes C. group for second stage. Third testing of both groups completes the process.3. Posttest only comparison group - less similar to experiments in that before-after testing and proper C. group are both lacking, but still an acceptable design in some situations.
There are several descriptive group-level designs presented in the text which are not quasi-experimental designs because they lack a control or comparison group. They are worth inclusion here as they are use for descriptive research in spite of their shortcomings.
4. Randomized one-group posttest-only design - this involves a good method of selecting elements but a poor system of controlling for extraneous variables during the research process.
5. One-group pretest-posttest design - this includes a good method of measuring a difference but no control or comparison group. It may be acceptable if there is a good deal of information about characteristic scores for the population.
6. Interrupted time-series design - this is similar to single-subject designs, but measures the dependent variable for a group rather than individuals.
