Imputation methods for doubly censored HIV data

Abstract

In medical research, it is common to have doubly censored survival data: origin time and event time are both subject to censoring. In this paper, we review simple and probability-based methods that are used to impute interval censored origin time and compare the performance of these methods through extensive simulations in the one-sample problem, two-sample problem and Cox regression model problem. The use of a bootstrap procedure for inference is demonstrated.

Keywords:

• bootstrap
• Cox regression model
• interval censoring
• Kaplan–Meier curve
• logrank test

Acknowledgements

This work was supported in part by a Dissertation Fellowship from University of Iowa Graduate College and an NIH/NIAID grant: R01 058740.