Quasi-experimental design is a research method that uses statistical controls as opposed to physical controls in an experiment. A statistical control is a variable used to test relationships between other variables in a social sciences study; it can also be called the control variable or the test variable. A difficulty in social sciences arises because human behavior is so complex that often we need to take into account three or more variables at a time, not just two variables as in the traditional scientific method of study. We call the need to test multiple variables multivariate analysis. Using a control variable in multivariate analysis allows us to see whether or not there is a relationship between variables, to see "how or why these variables are related," and to see if the relationship between variables holds for different people ("Chapter 3--Introducing a Control Variable (Multivariate Analysis)," Social Science Research & Instructional Center). The Dictionary of Sociology (1998) gives us the example of a study to find the relationship between unemployment and clinically diagnosed depression. A scholar might see that there may be a relationship between unemployment, depression, and social class as well; therefore, the scholar could use social class as the control variable by dividing the research data between working-class individuals and middle-class individuals. Results might show that, for all social classes, those who are depressed are also likely to be unemployed. Or, results may show that those of the working-class are no less likely to be depressed than those of the middle class; however, those of the working class have increased unemployment rates, and those who are unemployed have increased depression rates ("Statistical Control").
There are multiple quasi-experimental designs. One is the one-group pretest-posttest design. Using this method, one group of individuals is pretested, meaning tested before the actual experiment, on a dependent variable. A dependent variable is the variable that can be changed by the independent variable to see the relationship between the independent and dependent variable. In the case of participants of a support-group service, the participants would be considered the one group, whereas the support-group service would be considered the independent variable. Since the development of forgiveness is becoming accepted as an important healing tool for divorce support groups, one variable we might pre-test in a study of a divorce support group is levels of forgiveness. If all members of the group enter with low levels of willingness to forgive but develop higher levels as their group sessions continue, then we know it is the support group, as the independent variable, that is influencing the variable of willingness to forgive (Aysta, A. "A Quantitative Study of Forgiveness and Divorce Adjustment in Divorce Recovery Groups," Capella University). A post-test is a test of variables conducted after the experiment. In the above scenario, testing forgiveness levels after a number of support-group sessions would count as a post-test.
The one-group pretest-posttest design is useful as a quasi-experimental design because it allows for candidates to be tested both before and after treatment, which can better show the effect of the dependent variable; however, the method is still considered a weaker experimental design because it does not "control for potentially confounding extraneous variables such as history, maturation, testing, instrumentation, and regression of artifacts" ("Ch. 9: Experimental Research," University of South Alabama).
Ideally speaking, using random sampling in experiments is best for achieving reliable results; however, random sampling isn't always possible. As a consequence, scholars resort to using non-probability sampling in which samples are chosen based on what samples are available to the scholar and on the judgements of the scholar; judgements are based on what the scholar wants to achieve. Using non-probability sampling can be beneficial when the scholar does not plan to draw conclusions based on the results that can be generalized to the entire population. As an example, in the study above concerning the development of forgiveness in divorce support groups, using non-probability sampling would be very effective if the scholar wants to study what techniques are effectively being used by a particular support group.
No comments:
Post a Comment