A Score Test for Non-Informative Censoring Using Doubly Sampled Grouped Survival Data
提出一个封闭形式的得分统计量，用于检验非信息删失假设，适用于对删失个体随机子样本追踪确定失效时间的情形，并给出分层样本的扩展。
A closed form score statistic for testing the assumption of non-informative censoring is proposed for situations in which a random subsample of the censored subjects is followed to determine failure times. The test is derived for grouped survival data under the assumption of a proportional odds model for the hazard rates of censoring. Extensions to a stratified sample are provided. An example is given in which the possible association of discharge time with post-surgical infection rates was investigated by using data from an Israeli study of hernia patients. The test showed that longer hospital stays were associated with higher infection rates