A MULTIVARIATE MODEL FOR REPEATED FAILURE TIME MEASUREMENTS

A parametric multivariate failure time distribution is derived from a frailty-type model with a particular frailty distribution. It covers a s special cases certain distributions which have been used for multiva riate survival data in recent years. Some properties of the distributi on are derived: its marginal and conditional distributions lie within the parametric family, and association between the component variates can be positive or, to a limited extent, negative. The simple closed f orm of the survivor function is useful for right-censored data, as occ ur commonly in survival analysis, and for calculating uniform residual s. Also featured is the distribution of ratios of paired failure times . The model is applied to data from the literature.