Statistical Mediation and Moderation in Program evaluation: two key principles that evaluators should use when evaluating their programs.

Abstract Information: The core principle in program development and evaluation is the construction of interventions designed to impact recipients of the program's services. It is also widely recognized that the effects of these programs on participants can vary depending on individual attributes, such as sex, gender identity, ethnicity/race, and other relevant characteristics. These two principles are intimately connected with two key concepts within the domains of social, behavioral, and educational sciences, particularly in the statistical context: Mediation and Moderation. Nevertheless, program evaluators rarely use these concepts when analyzing and reporting the results of their evaluations. In this demonstration, we will go over the specific quantitative methods and provide specific examples of how an evaluator can test both mediation (i.e., the program is affecting the outcome in the way it was intended) and moderation (the program produces different outcomes based on participant’s characteristics such as sex, ethnicity, etc.). The demonstration will help participants: 1) define mediators and moderators in a program, 2) develop specific models that can be used to evaluate the effect of mediators and moderators, and 3) provide specific examples of how mediators and moderators can be tested using popular software tools like SPSS and R. The demonstration will also provide coding examples for implementing and testing these models in SPSS (through linear regression) and R (using both linear regression and path/structural equation modeling).