ABSTRACT
In this chapter, the authors explore some core methods, which they will build upon in subsequent multivariate failure time methods presentations. There are many variations in the criteria for study subject selection and follow-up with univariate failure time data. Other sampling variations include interval censorship, rather than right censorship only, and observation only of whether or not failure has occurred prior to an individual-specific assessment time. Many of these variations have a substantial corresponding statistical literature and are of considerable applied importance. Covariate measurement error is another important aspect of many univariate failure time data sets, particularly in such areas as nutritional or physical activity epidemiology where covariates of interest may be poorly measured. Naturally with large-scale clinical trials of this type there are many relevant analyses beyond those intention-to-treat comparisons. For example, some hazard ratios are decidedly variable over follow-up time within a particular study phase.
