What is the Quasi-Experimental Design?
"It is an experimental design that does not meet all the requirements necessary for controlling the influence of extraneous variables"
"If you want to make causal statements, you should conduct a true experiment. If you cannot use random assignment to design a true experiment, a quasi experiment is the second best choice. A quasi experiment is a fallback position, not as good as a true experiment far superior to pre-experiments or weak experiments"
"Nonetheless, random assignment of subjects to the experimental and control groups is sometimes not possible, particularly in field studies. Such experiments have been termed quasi-experiments by Campbell and Stanley to indicate that random assignment of subjects to treatment groups was not accomplished. Howver, the reseacher should be aware of the special problems that may arise when subjects are not assigned randomly to groups, and should take steps to solve them" (W. Borg & M. Borg, (1988). Educational Research: An Introduction, p.680).
"Quasi experiments are research designs that do not have randomly assigned treatment and comparison groups. Instead the comparisons between treatment and nontreatment conditions must always be made with nonequivalent groups or with the same subjects prior to treatment. If you cannot randomly assign people or groups to treatment conditions, you lose the ability to control what happens to whom. You can still observe what happens when, and to whom, however; and by deciding what and when to measure, you can design one of several quasi experiments. You forfeit control when you not use random assignment, but you can still conduct research and analyze cause-effect relationships without a true experiment" (L.H.Kidder, Research Methods in Social Relations, 1980, p. 43).
Quasi-Experimental Research Designs
1) Static-Group Comparison Design
X O
--------------------------------
O
Both the experimental and control group are given the posttest but NOT the pretest. It is the same as the "Posttest-only Contro-Group Design: discussed earlier except that there is no random assignment. What are the main threats to internal validity with this design?
- If there are differences in the posttest score, it may be attributed to the differences in the characteristics of the subjects in the group and the treatment. How does one differentiate which is which?
This design is a relatively weak design and the researcher should consider administering a pretest to both the groups to enable the making of stronger inferences.
2) NonEquivalent Control-Group Design
O1 X O2
--------------------------------------------------
O2 - O2
Note: The dotted line indicates that there is no random assignment of subjects to the groups.
This is the most common quasi-experimental design. It is the same as the Before-After Design discussed, except that subjects are not randomly assigned to the experimental and control group. Because of logistical reasons it is not possible to randomly assign subjects, so intact classes or all students in a particular class receive the treatment or act as the control group. What are the main threats to internal validity with this design?
- There is the possibility that group differences on the posttest are due to pre-existing differences rather than to the treatment effect.
To increase the strength of the design, Analysis of Covariance (ANCOVA), a statistical procedure is suggested. ANCOVA reduces the effects of initial group differences statistically by making compensating adjustments to the posttest means of the two groups.
O1 Reading Comprehension Method O1
Reading Ability ----------------------------------------------------------------------
Test O2 - O2
(covariate)
The experiment aims to establishe effectiveness of a "reading comprehension method" on comprehension of science texts. However, you are concerned about the initial differences in reading ability between the subjects in the two groups of which you cannot control becaus you are using intact classes. You could equate reading ability by giving a test before the experiment. The experimental and control group are equated on 'reading ability' by using it as a covariate in the ANCOVA. The experiment revealed that there was a significant difference between the experimental and control group means. This result indicates that the posttest means of the experimental and control groups are significantly different from each other, after taking into account pre-existing differences in reading ability.
"You should keep in mind that ANCOVA is an imperfect statistical technique for equating experimental groups prior to the treatment period. Only the variables that are measured can be used as covariates. The groups may differ on other variables, but if these variables have not been measured, they cannot be entered into the ANCOVA" (W. Borg & M. Borg, 1988, Educational Research: An Introduction, p.684).
