Event history graphs for censored survival data.

A compact graphical device for combining survival and time-varying covariate information is proposed. The proposed graph contains the Kaplan-Meier estimator for right-censored data and a simultaneous display of the behaviour of time-dependent covariate(s) and the lifetime for each subject in the sample. The observed levels of time-dependent covariates are possibly subjected to an initial dimension reduction or smoothing step to produce a continuous covariate function. Values of this function are plotted on a horizontal bar for the length of the lifetime of the subject. Covariate information for censored data is also incorporated. The union of the horizontal bars forms the Kaplan-Meier estimator of the survival function. Our graphical method is implemented with a new S-plus function and demonstrated in several applications.