Abstract: The purpose of this paper is to indicate that the time-dependent frailty model for multivariate counting processes of Chang and Hsiung (1996) can be applied to sequential analysis of paired survival data with staggered entry. It is shown that the efficient estimating function in calendar time is a martingale and, with a data-dependent change of time, it is asymptotically a Brownian motion process, which paves the way for sequential analysis. This approach is illustrated by re-examining the Freireich et al. (1963) data of remission lengths in leukemia. A simulation study is also included to indicate its performance numerically.
Key words and phrases: Multivariate counting processes, paired survival data, sequential analysis, time-dependent frailty.