The purpose of this proposal is to develop methodological and statistical procedures to determine how prevention programs change outcome variables. These mediation analyses assess the link between program effects on the constructs targeted by a prevention program and effects on outcomes. As noted by many researchers, mediation analyses identify the most effective program components and increase understanding of the underlying processes leading to-drug use. The need for these new procedures became apparent in the analyses of several NIDA funded drug prevention programs. The P.I. of this grant received a one-year NIDA small grant to study the assessment of mediation effects. In this proposal, we request four years of funding to develop methods and statistical procedures for the more complicated types of designs and procedures commonly used in prevention research. In the first of seven proposed studies, we examine type 1 error and statistical power of tests of mediation. In Study 2, mediated effects in multiple mediator and longitudinal models are examined. In Study 3, we develop and evaluate methods to estimate mediated effects in the nonequivalent control group design. In Study 4, we develop and evaluate methods to estimate mediated effects in logistic regression and the proportional hazards model. In Study 5, we develop and evaluate methods to estimate mediation effects in a variety of designs, including the hierarchical regression model and models with both moderator and mediator effects. In Study 6, we examine the effects of missing data, measurement error, non-normal data, and outliers on the estimation of mediated effects. In Study 7, we apply the new methods in several NIDA funded drug prevention projects. As part of Study 7, we will increase the use of mediation methods by the establishment of a consultation service and an interactive computer program to teach researchers how to conduct mediation analysis.
Showing the most recent 10 out of 85 publications
