Improving the Efficiency of Relative-Risk Estimation in Case-Cohort Studies

摘要: The case-cohort design is a common means of reducing the cost covariate measurements in large failure-time studies. Under this design, complete data are collected only on cases (i. e., subjects whose failure times uncensored) and subcohort randomly selected from whole cohort. In many applications, certain covariates readily measured all cohort members, surrogate expensive also may be available. existing relative-risk estimators for disregard outside sample thus incur loss efficiency. To make better use available data, we develop class weighted with general time-varying weights that related to proposed by Robins, Rotnitzky, Zhao. shown consistent asymptotically normal under appropriate conditions. We identify estimator within maximizes efficiency, numeri...