ABSTRACT
This chapter presents regression methods for marginal single outcome hazard rates and discusses the copula approach to higher dimensional failure time data. Baseline hazard models will be natural in some settings, for example in studies of failure time among littermates in animal experiments, or studies including multiple generation kindreds in genetic epidemiology. The chapter describes a more flexible modeling approach that focuses on regression associations for both marginal single, and marginal double, failure hazard rates using semiparametric Cox-type models. A simple modification allows double failure hazard rate models to be specified only for selected composite pairwise outcomes of interest, thereby somewhat reducing modeling assumptions, including that of independent censorship, to the set of modeled marginal single and double failure hazard rate processes. The applicability of an independent censoring assumption is an important consideration when applying marginal hazard rate models to correlated failure time data.
