关键词: 序贯概率比检验, 测试性验证, 序贯网图检验, 先验信息, Bayes小子样理论

Abstract: The current sequential probability ratio test method leads to the uncontrollable sample size of testability verification, and the sequential mesh test method can make the determined sample size be still large without using the prior information. A method for determining the sample size of testability verification is proposed, which is based on Bayes small sample theory and sequential mesh test. According to the prior distribution of testability indicators and the related parameter constraint values, the parameter space of indicators is divided based on the sequential mesh test method, and the Bayes factor and its threshold calculation method are given.Once the insertion position of a checkpoint is determined, the maximum sample size required for testability verification could be given. At the same time, a censoring strategy of sequential mesh test based on Bayesian small sample theory was designed, in which the consumer’s and producer’s risks are considered. The proposed method was validated by an example, and compared with classical testability verification method, sequential probability ratio test method, traditional sequential mesh test method and sequential posterior odds test method. The results show that the censored sample size and average sample size determined by the proposed method are better than those of other methods, and the consumer’s and producer’s risks can be effectively reduced. Key

Key words: sequentialprobabilityratiotest, testabilityverification, sequentialmeshtest, priorinformation, Bayessmallsampletheory