Semiparametric Likelihood Estimation in the Clayton–Oakes Failure Time Model (Q4956041)

When a sample consists of clusters and each cluster contains several possibly dependent failure times then multivariate failure time (MVFT) data arise. The analysis of such data must account for the potential dependence of failure times within the same cluster. One approach to modelling the intracluster dependence is through specification of parametric copula models. These models express the joint distribution of failure times as a function of the marginal distributions and a dependence parameter. Such a model is the Clayton-Oakes model [\textit{D.G. Clayton}, Biometrika 65, 141-51 (1978; Zbl 0394.92021); \textit{D. Oakes}, J. R. Stat. Soc., Ser. B 44, 414-422 (1982; Zbl 0503.62035)]. This paper considers estimation in the Clayton-Oakes model when the marginal hazards obey a \textit{D.R. Cox} [ibid. 34, 187-220 (1972; Zbl 0243.62041)] hazards model. The marginal Cox specification within a Clayton-Oakes model for dependence combines marginal hazard ratio parameters for covariate effects with a single, interpretable parameter for intracluster dependence. Parameter estimation is based on an approximate generalized maximum likelihood estimator. A model application with example datasets is illustrated.