Joint Modeling of Longitudinal and Survival Data with Censored Single-index Varying Coefficient Models

Abstract: In medical and biological research, longitudinal data and survival data types are commonly seen. Traditional statistical models mostly consider to deal with either of the data types, such as linear mixed models for longitudinal data, and the Cox models for survival data, while they do not adjust the association between these two different data types. It is desirable to have a joint modeling approach which accomadates both data types and the dependency between them. In this paper, we extend traditional single-index models to a new joint modeling approach, by replacing the single-index component to a varying coefficient component to deal with longitudinal outcomes, and accomadate the random censoring problem in survival analysis by nonparametric synthetic data regression for the link function. Numerical experiments are conducted to evaluate the finite sample performance.