Testing for Sufficient Follow-Up and Outliers in Survival Data
本文开发了一种诊断检验，用于判断生存数据中是否进行了充分随访，并基于犯罪学数据提出了异常值检验方法，适用于存在免疫个体的情形。
Abstract A situation in which a sample of failure times with many of the largest times censored may be taken as evidence of a proportion of "immune" or "cured" individuals, who are not subject to failure, in the population. A plot of the Kaplan-Meier empirical distribution function for such data will tend to level off near its right extreme, provided that follow-up of individuals has been continued for long enough. This article develops a diagnostic test for sufficient follow-up in samples where there may or may not be immunes and demonstrates its properties. The procedure is illustrated on some criminological data, leading us also to propose a test for outliers in this kind of situation. Key Words: Censored failure dataFollow-up timeImmune proportionKaplan-Meier estimatorOutliers in Weibull or Exponential dataRecidivism