ABSTRACT
Survival analysis involves a large collection of statistical methods to deal with survival data introduced in the previous chapter. Statisticians are usually interested in studying the unknown numeric characteristics of a population, or the so-called parameters. To achieve a close and meaningful “guess” of the unknown parameters, we may have to consider some computational methods whose theoretical properties are justifiable. The process of acquiring such a meaningful “guess” is called an estimation. In survival analysis, it is often impossible to simplify the estimation task to a couple of distribution parameters, and we are interested in determining the complete distribution of survival time. The objective is thus not a few parameters, and we have to provide the functional estimate of the whole survival distribution curve at all time points. The estimation method for this kind of problem is called nonparametric estimation, since the estimand is not a traditional finite-dimensional “parameter.” The nonparametric estimator will be introduced in Section 2.1.
