The frailty model. (Q2468953): Difference between revisions

Clustered survival data are encountered in many scientific disciplines including human and veterinary medicine, biology, epidemiology, public health and demography. Frailty models provide a powerful tool to analyze clustered survival data. This book introduces the basics of frailty model methodology for survival data, including parametric and semi-parametric frailty models and accelerated failure time models. Real-life data are used to demonstrate how particular frailty models can be fitted and how the results should be interpreted. The programs to fit all the worked-out examples in the book are available from the Springer website with most of the programs developed in the freeware package Winbugs and R, a language and environment for statistical computing. The book starts with a brief overview of some basic concepts in classical survival analysis, collecting what is needed for the reading on the more complex frailty models. Then the parametric proportional hazards models with gamma frailty are introduced, and a Bayesian approach is briefly discussed. The scope is then widened to alternatives to the simple frailty model, especially fixed effects models, stratified models, and what the authors call the marginal model, a model where the clustering is ignored, but standard errors are modified to accommodate for the clustering. Alternatives to the gamma frailty distribution are discussed, for instance the inverse Gaussian, compound Poisson, and log-normal distributions. Semi-parametric models are studied, and it is pointed out that the estimation of such models is more difficult than in the fully parametric setting. Several approaches, including the EM algorithm, penalized partial likelihood, and the MCMC algorithm, are compared. Then multi-level models, i.e., models with more than one frailty term, are discussed. This includes both random slopes and random intercepts in the same model, and hierarchical models, where one clustering level is nested within another clustering level. Finally, further extensions of the frailty model are discussed: Correlated frailty models and accelerated failure time frailty models.