Estimation in a general semiparametric hazards regression model with missing covariates

Abstract

In survival analysis, missing observations are often encountered in covariate measurements, and ignoring this feature may make an invalid inference. In this article, we consider a general semiparametric hazards regression model for right-censored data with some covariates missing at random. The covariate effects in this model are characterized by a time-scale change and a relative hazard ratio. A class of weighted estimators are proposed, and the resulting estimators are shown to be consistent and asymptotically normal. Furthermore, fully augmented weighted estimators are also studied to improve estimation efficiency. Simulation studies demonstrate that the proposed estimators perform well in a finite sample. An application to the mouse leukemia data is provided.

Keywords:

• General hazards regression
• missing at random
• relative hazards ratio
• time-scale change
• weighted estimating equations

Acknowledgements

The authors would like to thank the Editor, Professor N. Balakrishnan, an associate editor and the reviewer for their constructive and insightful comments and suggestions that greatly improved the paper.

Additional information

Funding

This research was supported by the National Natural Science Foundation of China (Grant Nos. 11771431, 11690015 and 11926341) and Key Laboratory of RCSDS, CAS (No. 2008DP173182).