|
Abstract:
|
Subgroup analysis has recently been received much attention in many fields of study, such as clinical trials and policy marketing. In this paper, we propose a score type test statistics for subgroup identification in the cox PH model with censored survival data. The proposed test statistics is robust against the misspecification of the baseline effect model and when the null hypothesis of non-subgroup effect is rejected, the subgroup effect can be easily estimated based on the test statistics. We derive the asymptotic distributions of the test statistics under the null and local alternative hypotheses. The sample size required for the test to obtain sufficient power is also determined.
|
