Regression Discontinuity: A Design Solution

Although experimental designs are considered the ideal for social work researchers conducting intervention research, they are not feasible or ethical in many study situations. Regression discontinuity (RD) designs, which are considered the next best design after experiments by Cook, Shadish, and Campbell (2002, 2008) and Trochim (2010), have a number of desirable characteristics. First, they allow researchers to target the cases that are most in need of services, avoiding one of the major ethical barriers to many experimental studies. Second, they remove the burden of establishing that the treatment and control/comparison groups in an intervention study are equivalent. According to Cook and Shadish (2008), these advantages are based in the fact that in a well-executed regression continuity design, the differences in the treatment and control group are known and measured. It is when group differences are unknown that they threaten internal validity and are problematic. Content. This workshop will provide an overview of the issue of counterfactuals in experimental and quasi-experimental designs as a foundation. Then, the RD design and its associated assumptions and requirements will be discussed. Information on how to carry out a RD design and analyze and interpret data from such a design will also be covered. Presentation plan and pedagogical techniques: 1. An introduction to the major concepts (presentation), 2. Interactive discussion of participants' experiences meeting demands of experimental designs and how RD might apply to their research situations, 3. Presentation and discussion of an example of an RD design in school research, 4. Demonstration of analysis techniques in SPSS and STATA, 5. Practice interpreting analysis results, 6. Concluding Q and A. Participants will be provided with a detailed handout containing the Powerpoint, instructions for conducting a RD study, syntax for analyzing data, and annotated results. Audience. This workshop will be suitable for researchers and PhD students will all levels of experience. Knowledge of basic regression analysis will be helpful, but is not required. Conclusion. Cook, Shadish, and Campbell (2002) highly recommend RD designs and think they should be used more often. This workshop will help social work researchers determine if RD solves their experimental design woes and, if so, how to carry out an RD study from the planning to analysis stage.