A psychology professor is interested in whether implementing weekly quizzes improves student learning. She decides to use the weekly quizzes in one section of her introductory psychology class and not to use them in another section of the same course. Which type of quasi-experimental design do you recommend for this study???
The professor wants to determine whether using weekly quizzes improves student learning. This can be done using a non-equivalent groups design. Here, the professor divides her psychology class into two sections, students in one of the section are given quizzes every week and the performance of the students is evaluated in the quizzes. The other group of students are not given weekly quizzes. At the end of the course, the performance of the students of both the groups would have to be evaluated using a common test. For the results of this experiment to be meaningful it is essential for the professor to ensure that when the students are divided into the two sections both the groups have the same level of prior academic performance. Also, the way the two groups are taught should be the same. For an accurate result, the only thing that should be different for the two groups is that one is made to take the weekly quizzes and the other is not. There should not be any other variations.
This kind of design would involve a control group (the class without the quizzes) and a treatment group (the group with the quizzes). In this design, you would assign a pretest to gather data from both groups before implementing the quizzes, then a post-test on both groups after you have finished teaching the lesson. A possible hypothesis could be that the quizzed group would do better on the post-test because they have had more interaction with the material through assessment.
There are a number of quasi-experimental designs, including the difference in differences, non-equivalent control groups, regression discontinuity, case-controlled, and panel analysis designs.
The regression discontinuity design can find differences near the cutoff from a nonrandomized treatment group and the control group. (e.g. scholarships given to the top 20% of students -- regression continuity looks at the performance of people near the 80% mark -- above and below -- to see if there is a difference.)
A case controlled design typically looks at the results, and then looks at the groups to find a "causal" link.
A good choice is the difference of differences design. Here a pretest is given, the groups are assigned to be control or treatment groups, and a post test is given. It is assumed that the change in the control group is the naturally occurring change, and any other change in the treatment group can be assigned to the treatment.