摘要
本文从Bayes观点出发,利用Kullback-Leibler距离和微分几何方法,系统地讨论了微小扰动对于Bayes统计推断的局部影响,给出了Bayes局部影响分析的一般公式作为应用,具体讨论了线性回归模型的Bayes估计和Bayes预测的局部影响问题,数值计算的结果表明,本文的方法是比较有效的。
A general method is devleoped in this paper for assessing local influence of minor perturbations of a sratistieal model in Bayesian analysis[1-3] .The method is based on the Kullback-Leibler divergence and cersain elementary ideas from differ-ential geometry .ami it seems to providen muified approach for handling a variety of diagnostics problems in Bayesian nalysis. In appllcations.the local estimative influence and local predictive influence in linear regressioa are discussed from Bayesian point of view. Some numerical examples art; considered ind compared with the results of [4].
出处
《应用数学》
CSCD
北大核心
1995年第2期237-245,共9页
Mathematica Applicata
基金
The projict is supported the National Natural Science Poundation of China.
关键词
贝叶斯分析
线性回归
贝叶斯估计
贝叶斯预测
Baycsian analysis
Curvature
Influence
Kallback-Leibler divergence
Predictive distribution
