A Unified Treatment of Locally Most Powerful Rank Tests Under Type II Censoring
提出一种通用方法，统一处理多种检验问题和II型删失方案下的局部最优秩检验，简化了传统推导，并应用于之前仅有特设检验的问题。
Abstract For the derivation of locally most powerful (LMP) rank tests, a general method that encompasses a wide variety of testing problems and type II censoring schemes is formulated. Applications are illustrated in the context of the one-sample, two-sample, and regression problems. It is shown that the traditional derivations of the LMP rank tests are greatly simplified with the proposed method. The LMP criterion is also applied to some problems for which only ad hoc tests were previously proposed.