Risk in Context: New Methodology for Modeling Risk by Context Interactions

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Theoretical work on how contextual environments affect health outcomes has long stressed the importance of interactive effects: contexts are often expected to affect individuals differently. In fact, risk for poor developmental outcomes is often conceived of as an interaction - risk factors may have negative effects only for those experiencing high levels of stress or multiple risk factors. Research examining these interactions, however, has typically been limited to studying simple bivariate interactions. The purpose of this study is to develop and demonstrate the utility of a new statistical tool, random effect regression mixture models, to further the study of risk in context. These models work by identifying latent classes of individuals who are differentially affected by their contexts and then provide the tools to understand what differentiates these children. The study will evaluate the conditions under which valid inferences can be made using these models and then demonstrates their use to assess differential effects of stressful life events. These models will then be further developed to allow their use with multilevel data through the inclusion of random effects. Monte Carlo simulations will be conducted to evaluate the utility of the models in conjunction with analyses conducted on a large sample of high-risk children to demonstrate the utility of the models with real data. The results will focus on demonstrating the validity of this model when used to study how developmental contexts affect children differently. In a final set of analyses multilevel random effects models will be applied to data from a second study to assess the effects of stressful life events on the development of depression. This application is important for providing evidence that previous results are not situation specific. Further, as the second study was designed to look at complex context by risk interactions this application should demonstrate the power of these models as a tool for answering real world reseach questions.