ABSTRACT
This chapter provides additional attention to an important class of multivariate failure time data analysis contexts where individual study subjects may continue follow-up beyond a first failure to second and subsequent failures on the same time axis. Moreover, such failures may be of the same or different types. Cox-type regression models were applied to the data, with baseline rates stratified on age at enrollment and randomization status in the companion Women’s Health Initiative hormone therapy trial. Marginal modeling and intensity process modeling provide a two-pronged approach to the elucidation of failure rate and regression associations with recurrent event data. Marginal models for single and double failure hazard rates can offer a rather direct assessment of the relationship of treatments or exposures encoded in a covariate history to readily interpretable event rates. The majority of applications will likely benefit from the inclusion of analyses of marginal single failure hazard rate models or related mean models, with recurrent event data.
