论文部分内容阅读
分析了基于Jeffreys验前的经典Bayes方差估计以及考虑验前信息可信度情况下Bayes方差估计存在的问题,在一般情况下,其方差估计要大于验前子样和验后子样的方差,这显然是不合理的.这是采用Jeffreys验前和正态共轭分布假设时存在的固有问题.为了解决这一问题,提出了方差估计的修正公式,经过计算验证,其值在验前子样和验后子样方差之间,说明修正公式是合理的.
This paper analyzes the problems of Bayesian variance estimation based on Jeffreys pre-test and Bayes variance estimation under the condition of prior information credibility. In general, the variance estimation is larger than the variance of pre-test and post-test sub- This is obviously unreasonable.This is an inherent problem when using pre-Jeffreys pre-normal and normal-conjugate distribution hypothesis.In order to solve this problem, a modified formula of variance estimation is proposed, after calculation and verification, Sample and post-test sub-sample variance, indicating that the correction formula is reasonable.