基于异常值检验的Bayesian方法验前信息可信度计算

Prior Information Creditability Calculation of Bayesian Method Based on Outliers Test

由于采用了验前信息,Bayesian方法为小子样条件下的装备试验与评价提供了可行的理论依据,从而在装备试验与评价领域得到了广泛的应用。然而,由于验前信息具有多源性及主观性,当采用其进行Bayesian统计推断时,可能会导致Bayesian统计推断的结果偏差较大,使得决策风险增加。因此,研究Bayesian方法中验前信息的可信度显得至关重要。在现场试验样本信息已知的前提下,给出了验前信息可信度的定义,提出一种基于异常值检验的验前信息可信度计算方法,并基于Matlab数学软件,通过仿真思路实现了算法求解。最后,以某雷达探测距离试验为例,证明方法具有科学合理性。

Using the prior information, Bayesian method provided a feasible theoretic basis for small-sample teat and was widely used in the area of equipment test. Nevertheless, using prior information to make Bayesian sta- tistical inference may lead to large deviation and high risk due to the multi-source and subjectivity of prior informa- tion. Therefore, it is important to do the research in prior information creditability of Bayesian method. Under the known field testing information, a simulation method of prior information creditability based on outliers test and re- alizes the algorithm by Matlab is proposed. Finally, the rationality of the method is proved by an equipment testing instance.