|
Abstract:
|
Abstract Mediation analyses play an important role in exploring how independent variables affect dependent variables through intermediate variables or called mediators. There has been a growing demand for mediation analyses for high dimensional data , specifically for high-dimensional epigenetic data, where the number of potential mediators is more than half a million. While existing statistical approaches conduct mediation analyses for single or multiple mediator models, none of these methods deal with high dimensional mediators while controlling for both Type I and Type II errors. Even more specifically, for repeated outcomes, which are often observed in longitudinal studies, there are no software or packages to perform an efficient screening . In this body of study, we developed a novel screening method, emscreening, to perform a screening process for high dimensional mediators with a repeated outcome. Simulation studies were used to evaluate the performance of the proposed joint screening method. For both continuous and binary outcomes, the proposed joint screening method showed comparable sensitivity and specificity when the number of mediators in the model was relatively s
|
