|
Abstract:
|
In survival analysis a frailty is an unobserved random effect. Usually this is taken to act multiplicatively on the hazard function. Introduction of frailties into univariate survival models typically has the effect of attenuating effects of fixed covariates. In multivariate survival analysis, introduction of a common frailty, or of correlated frailties, among related individuals is a useful approach to modeling dependence between their survival times. Features of the frailty distribution are reflected in dependence characteristics of the observable joint survival distribution. There is a very natural connection with copula models and with nonparametric measures of dependence. The talk will discuss these connections and certain properties, such as invariance under censoring and truncation, that are important in biostatistical applications of copula models.
|
