ABSTRACT
This chapter discusses the distribution-free approach in the proportional hazards context and assume no parametric form for the baseline hazard. The foundation work in this area was done by Cox. This work has become a platform for building the methodology of the last 30 years. The Cox model is the most important distribution-free regression model used for the analysis of censored data. Central to the method of Cox is the idea that the hazard function is the basis of the regression model. Essentially, the censoring of an individual has the simple effect that there is one less individual 'at risk' at the censoring time, so that censored individuals drop from the risk sets as time passes. An extensive discussion of model fitting is given in Collett. Included in this discussion are the standard regression techniques for using dummy variable covariates to define a factor model for incorporating a factor A with a levels.
