二项分布中成功概率的贝叶斯序贯检验方法

针对成败型效应实验研究中二项分布未知参数(成功概率)的假设检验问题,在复杂假设条件下,提出了一种基于Bayesian验后概率的序贯检验方法,建立了检验的判别准则,给出了判别准则临界值的计算方法.在给定截尾实验次数的条件下,提出了一种截尾方案,建立了截尾判断方法.最后结合示例,对上述方法的应用过程进行了说明,并和现有方法进行了分析比较.结果表明:在一定先验信息的条件下,该方法给出的检验样本量远小于经典方法确定的检验样本量.

Bayesian sequential test method for probability of success of binomial distribution based on posterior probability

A sequential test method based on Bayesian posterior probability was proposed to perform tests of hypotheses about probability of success of binomial distribution in complicated hypothesis conditions. The decision rules were constructed and the method for calculating the criterion value was given. For a given maximal sample size, the truncated method of the judge criteria was discussed for hypothesis test. At last, an example illustrated the proposed method and the result was compared with classical methods. The result shows that the sample number determined by the proposed method can be much smaller than that determined by classical methods on the condition of given prior probability.