This seminar discusses and demonstrates regression-based analytic strategies for detecting moderation and mediation. A moderation effect is also known as an interaction and arises when one hypotheses that the strength of the relationship between two variables 'depends upon' a third variable. Envision a model with X1 and X2 interacting in the prediction of Y. The regression model will be Y' = a + B1X1 + B2X2 + B3X1X2 + e. Strategies for overcoming the strong possibility of muliti-collinearity in interactions will be offered and examples will be demonstrated using SPSS. A mediation effect refers to a third variable 'intervening' in the relationship between two other variables such that the third variable serves as a generating mechanism between the other two variables. If causality can be realistically inferred, then in mediation, X causes M which in turn causes Y. Testing for mediation within the framework provided by Baron and Kenny (1986) requires three different regression equations, one of which is hierarchical. Sobel tests for the significance of the mediator will be discussed and regression-based tests for mediation in SPSS will be demonstrated. This is NOT a software seminar and is not presented in a computer lab. However, intermediate-level experience with multiple regression is required and beginning-level experience with SPSS is helpful (although not essential).
