小子样条件下正态先验信息的融合

Normal Prior Information Fusion Method Under the Small Sample

摘要:

对于Bayes小子样检验方法,先验信息的有效性影响着统计推断的准确性。在对多源正态先验信息进行 融合时,采用自助法将之转换为先验子分布,其中采用了改进的经验分布函数法进行再抽样。然后,基于专家水平 确定其权重,再结合专家估计法确定各先验子分布的权重。最后,通过计算综合先验分布与假设的正态分布的差 异部分的面积,对综合先验分布为正态分布的假设进行检验。案例分析表明此融合方法是有效的。

Abstract:

For the small sample Bayes verification method, the effectiveness of prior information affects the correct statisti? cal inference. As for the fusion of multi-source normal prior information, the bootstrap method was used to convert a piece of prior information into a prior sub-distribution. Then experts’weight was determined based on their level, and the weight of the prior sub-distribution was determined by expert estimate method. At last, by calculating the discrepancy pro? portion between the integrated prior distribution and the hypothetical normal distribution, the hypothesis that the fused in? tegrated prior distribution was normal distribution was tested, and the method to accurately calculating the distribution pa? rameters was also put forward. The analysis by an example showed the fusion model was effective and accurate.