Sankhya: The Indian Journal of Statistics
2001, Volume 63, Series A, Pt. 1, pp. 1--14
BAYESIAN ANALYSIS OF INTERVAL-CENSORED SURVIVAL DATA USING PENALIZED LIKELIHOOD
SUJIT K. GHOSH, North Carolina State University, Raleigh, USA
DEBAJYOTI SINHA, University of New Hampshire, Durham, USA
SUMMARY. Grouped survival data with possible interval censoring arise in a variety of settings. This paper presents a penalized likelihood method (also called posterior likelihood by Leonard, 1978) for the analysis of such interval-censored survival data. A penalty function based on the motivation from the auto-correlated prior process is used to incorporate the available prior information on smoothness of the hazard. A version of the EM algorithm (Dempster it et al., 1977) for the maximization of the penalized likelihood is used to obtain smooth estimates of the hazard function. We also discuss different methods to estimate the hyperparameter of smoothing. The methodology developed in this article is exemplified with the data for the times to cosmetic deterioration of breast cancer patients.
AMS (1991) subject classification. 62C10.
Key words and phrases. EM algorithm, OSL algorithm, prior likelihood.
Full paper (PDF)