Sequential Selection of the Larger of Two Normal Means
研究了在临床试验等场景中，如何通过序贯检验判断两个正态分布均值哪个更大，同时减少对较小均值总体的观测次数，提出了未知方差情形的扩展方法并验证其渐近最优性。
Abstract Sequential tests for deciding which of two normal means is larger are considered in situations, such as clinical trials, where it is desirable to reduce the number of observations made on the population with the smaller mean. We propose and study an extension to unknown variances of a test proposed by Flehinger et al. (1972) for known variances. This involves replacing the unknown variances by estimates and replacing parallel stopping boundaries (-b, b) by moving boundaries that decrease to (-b, b). The proposed test is shown to be asymptotically optimal, and simulation results show that it performs satisfactorily for small sample sizes.