J. de Uña Álvarez, A. Panduro Martín
The analysis of censored gap times is of interest in many studies but there are difficulties in the efficient estimation from such data using nonparametric or semiparametric methods. When the follow-up is limited and the times for a given individual are not independent, induced dependent censoring arises for the second and subsequent times. This invalidates the direct estimation of the marginal distributions through the Kaplan-Meier technique. We analize the copula model proposed by Lawless and Yilmaz (2011), characterized by a parametric copula model and nonparametric models of two marginal functions. We propose a sieve maximum likelihood estimation procedure, using B-splines, for estimating the marginal distributions and the copula parameter. We discuss the consistency and efficiency of the proposed procedure; these are novel results in the referred setup. The finite-sample performance of the proposed estimators is analysed in a simulation study. A real data illustration is given.
Keywords: Copula Functions, Maximum Likelihood Estimation, Multi-state Models, Semiparametric Efficiency, Sieve Estimation, Survival Analysis
Scheduled
GT18 Non-parametric statistics II. Nonparametric inference for density and distribution
June 9, 2022 12:00 PM
A15
