Session: Abstract of Distinction: Leveraging Modern Mediation Analysis in Prevention Science (Society for Prevention Research 26th Annual Meeting)

SESSION INTRODUCTION: In prevention science, it is crucial to study mediation in order to understand how an intervention can change people’s behaviors. The goal of this organized group symposium is to present current issues in mediation analysis to optimize research in prevention science. The theme is consistent to the SPR’s goal and 2018 conference theme to optimize prevention science by introducing innovative methods and statistics. Presenters will be coming from different institutions all over the world to present their individual and collaborative research on mediation analysis in prevention science.

The first paper, “Natural Indirect Effects in Survival Mediation Analysis” introduces a statistical model for mediation with a survival mediator and derives a natural indirect effect using the potential outcomes approach. An illustrative example (simulated data) is given in context of a preventive-intervention study.

The second paper, “Comparison and Application of Traditional and Potential Outcomes Framework Methods for Estimating Mediated Effects in a Three-wave Mediation Model” examines how well the potential outcomes approach in estimating mediated effects in a longitudinal mediation model performs compared to traditional approaches (ANCOVA, difference scores, & residualized change scores). A Monte Carlo study and application to ATLAS (Goldberg et. al., 1996) dataset is presented.

The third paper, “Bayesian Mediation Analysis with Latent Variables at Small Sample Sizes” presents Bayesian methods for mediation analysis with latent variables. A comparison of frequentist methods and Bayesian methods is made by a Monte Carlo study with N=50.

The fourth paper, “Person-oriented Approaches to Mediation Analysis” focuses on person-oriented statistical methods to study individual differences in the mediational process. A simulation study compares a person-oriented method with variable-oriented methods.

The fifth paper, “Impact of Reducing Assessment Length on Statistical Mediation Conclusions” studies the effect on mediated effects when reducing the number of items for a mediator. Results from a Monte Carlo simulation study is discussed.

It is expected that the modern approaches on mediation analysis will be appealing to the audience of the 2018 SPR Annual Meeting. At the end of the session, the discussant will make some summary statements and moderate a discussion between the attendees and the presenters.