More effective estimation for additive hazards model in generalized case-cohort study

Abstract

The generalized case-cohort design is a common way to save cost when the failure is nonrare event and some covariate measurements are expensive in large cohort study. Under this design, complete covariate data consists of a subcohort randomly selected from the whole cohort and a subset randomly selected from the remaining failures. To make better use of the available data, a series of weighted pseudo-score estimators are proposed in the additive hazards model. In this paper, we use covariate grouping auxiliary information to improve the estimation efficiency of the model. Through two different weight coefficients, the auxiliary information are transferred into different estimation equations. Generalized moment estimation is used to integrate information and estimate model parameters. If subjects registered in clinical studies are not the representative samples of the population with overall survival information sources, the generalized moment estimation is extended with the aim of more robust estimate. These estimators are all shown to be consistent and asymptotically normal under appropriate conditions. Both the theories and simulation studies conclude that our estimators are more efficient than the weighted pseudo-score estimator. We apply the proposed method to a real data.

Keywords:

• Auxiliary information
• Estimating equation
• Generalized case-cohort
• Generalized moment method