作　　者： (胡红波）; (孙桥）; (杜磊）; (范哲）; (白杰）;

机构地区： 中国计量科学研究院,北京100029

出　　处： 《计量学报》 2017年第5期656-660,共5页

摘　　要： 比较了GUM系列文件与贝叶斯方法评估不确定度的过程,说明GUM是以测量方程为基础的前向分析方法,而贝叶斯分析是以观测方程为基础、以数据分析为主的一种后向不确定度评估方法。运用概率模型的描述方法对上述两种分析过程进行了分析与比较,说明了两种方法在线性测量模型的条件下,不考虑被测量先验分布时两种结果基本一致,对于非线性的测量模型,GUM S1得到的结果与一定先验分布条件下的贝叶斯分析的结果也基本一致。最后通过实例说明了该结论。 The process of uncertainty evaluation methods based on GUM serials and Bayesian analysis is compared, and it is derived that GUM uncertainty evaluation is based on measurement equation, and a forward uncertainty evaluation method, while Bayesian analysis is an inverse uncertainty evaluation based on observation equation. A probabilistic approach to analysis and comparison is used to the two approaches. It＇ s concluded that for linear model and with noninformative prior for the measurand both analysis can be applied and get the same result, but for non-linear model, only GUM S1 method and Bayesian analysis give the same result. The results are illustrated by examples.

关 键 词： 计量学 不确定度评估 观测方程 贝叶斯分析