Estimation and inference in nonparametric Cox-models: time transformation methods (Q1887217)

The authors demonstrate how survival time data can be analyzed in a nonparametric manner. They deal with a generalization of the Cox proportional hazard regression model to a completely nonparametric model and focus on related estimation problems. They discuss the so-called time transformation approach based on transforming the estimator of the covariate function into an exponential regression problem via certain iterative algorithms. In the class of time transformation methods a new so-called covariate order method, which is based on arranging of the observations in increasing order of covariates, is particularly studied. It turns out that this method has a number of useful properties. Some suggestions about inference and model checking in the nonparametric Cox set up are discussed, and a brief overview of other methods and some practical issues are presented. Results of a simulation study and real data examples are analyzed.