ABSTRACT
The parametric methods described in Chapter 3 can be useful in the analysis of survival data when the survival time can be assumed to follow a certain distribution. If the distribution is correctly assumed, the resulting estimates are most efficient. However, incorrect specification of the distribution can lead to misleading conclusions. Semiparametric models without having to specify the distribution of the survival time are desirable when it is unclear which distribution should be assumed. In this chapter, we will discuss the inference associated with semiparametric Cox
regression model. We will review the partial likelihood used for estimation of the regression parameters and discuss issues associated with tied data, time-dependent
covariates, and model diagnostics. We will also review briefly on the extension of the Cox regression model to correlated events. Examples from biomedical studies will be used for illustration.
