Statistical Methods for Genetic Epidemiology

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Cost-effective cancer prevention and control strategies that emphasize population subgroups at highest risk are increasingly needed to counter national and global health costs. The proposed research will address this need by helping identify high-risk individuals in three specific aims. The first aim is to develop more informative measures of model performance that can focus on specific population subgroups. In particular, we will develop and evaluate better measures of model accuracy (calibration), and model discrimination between those likely and unlikely to develop the adverse outcome. The second aim is to develop new ways to expand the utility of epidemiologic data for assessing model performance. These methods will allow investigators to accommodate cohort selection bias in assessing model performance; use and interpret case-control data for assessing model discrimination; and assess risks of multiple competing adverse outcomes in the same individual. The third aim is to augment the freely available R-based software RMAP (Risk Model Assessment Program) http://www.stanford.edu/~ggong/rmap/index.html to include the new and more informative performance measures and allow investigators to apply them to a broad range of epidemiologic data.