Abstract

Recurrent event analysis has been a topic that has generated much attention in the last few years. The analysis of recurrent event data can be a complex task, with many different relationships and dependencies that need to be specified. One aspect of recurrent events that is difficult to specify, and even more complex to justify, is the relationship a subject’s rate of failure will have as events continue to occur. Marginal modeling is a method that estimates parameters by considering the marginal distributions of the data. In the thinking that the dependence between the events is not an interesting aspect of the analysis, marginal modeling ignores the dependence on event number for estimation of parameters and corrects for this dependence in the variance. The fundamental issues of marginal modeling in recurrent event analysis will be discussed, specifically when used in Cox proportional hazards regression. 1

Citations