Stratifying Subjects for Treatment Selection with Censored Event Time Data from a Comparative Study

摘要: The conventional approach to comparing a new treatment with standard therapy is often based on summary measure for the difference over entire study population. A positive trial respect such global measure, however, does not mean that all individual future patients would benefit from treatment. On other hand, negative finding may be sufficiently conclusive claim entirely futile. In this article, we propose systematic identify who an event time outcome via two-stage inference procedure. We first develop scoring index stratify parametric or semiparametric survival models observed times and covariates. then use nonparametric method estimate average each stratum defined by score. Sampling variation of resulting estimator also provided across spectrum score controlling certain local error rates. With numerical study, show proposal performs well under various practical settings. Our illustrated data recent clinical evaluate whether specific anti-hypertensive drug prolong lives stable coronary artery disease normal slightly reduced left ventricular function.