Statistical mediation is the extent to which the relation between a predictor X (often a prevention of treatment program) and an outcome Y is explained by a third variable, a mediator M, which transmits the influence of X to Y. Many data sets in the prevention sciences include variables that are hypothesized mediators, or mechanisms that influence behaviors. The goal of this symposium is to highlight recent advancements in mediation methods using applied prevention data.
The first poster in this session describes the process of conducting mediation for substance use outcomes, which are often zero-inflated. The author describes the unique challenges of computing mediated effects from models that include a zero-inflated component, and provides accessible formulas for mediated effects with a variety of models that handle zero-inflated count outcomes.
The second poster in this session uses a bi-factor model to examine the mediation role of coping on a behavioral intervention program. The authors used a bi-factor model and examined coping strategies as a mediator of later behavioral problems, and found that the bi-factor model was a good model for the mediator and measured different types of coping strategies as a mediator.
The third poster in this session combines mediation analysis with an optimization strategy to assess effectiveness of programs designed to reduce the gender achievement gap in STEM. The authors used this new combined method in a fully factorial design to determine whether perceived stereotype threat was a significant mediator of the program effect on task performance.
The fourth poster in this session compares the causal potential outcomes framework of mediation with traditional mediation analysis via Monte Carlo simulation, and demonstrates the methods with an applied example using smoking prevention data. The authors found that using the potential outcomes framework results in lower power than traditional mediation when all variables are binary.
In conclusion, the posters in this session complement each other by demonstrating various types of mediation models with data from prevention and intervention studies. Mediation continues to be an important method for examining mechanisms of change in prevention research.