成败型试验的Bayes序贯网图检验法

Bayes序贯检验法和序贯网图检验法分别从先验信息利用和对检验问题拆分的角度对序贯检验法进行了有效的改进。融合二者的优点,针对成功率检验问题,提出了一种Bayes序贯网图检验法,对该方法中先验信息的利用以及插入点的选择等都进行了详细的讨论。同时也给出了相应截尾方案及比较计算实例,并通过Monte Carlo方法计算了平均样本量。实验结果表明,该方法对序贯网图法的改进是全方位的,不仅可以降低二类风险,而且所需的截尾样本量和平均试验量也更少。

Bayesian sequential mesh test for success ratio

Bayesian sequential test and sequential mesh test provide ideas form two different ways to improve the sequential probability ratio test(SPRT) by using the prior information and dividing the original SPRT test.Taking advantages of the two methods,a new method,Bayesian sequential mesh test,is proposed for the test of success ratio,and is also discussed in detail about how to use the prior information and how to insert the new test points.Examples for comparing the truncated sample times and the average sample times are also given.The computational results show that the proposed method improves the sequential mesh test method,both by depressing the risk and by reducing the truncated sample size and the average sample size.