曲靖师范学院学报 ›› 2018, Vol. 37 ›› Issue (6): 1-4.

• 数学研究 •    下一篇

异分布混合模型及其参数的贝叶斯估计

张波1,刘鹤飞2,王坤2   

  1. 1.云南大学 党委研究生工作部,云南 昆明 650504;
    2.曲靖师范学院 数学与统计学院,云南 曲靖 655011
  • 收稿日期:2018-09-11 出版日期:2018-11-26
  • 作者简介:张 波,云南大学党委研究生工作部副教授,博士,主要从事数理统计研究.
  • 基金资助:
    曲靖师范学院校级青年项目“隐马尔科夫多维分布的参数估计”(2018QN004).

Bayesian Estimation of Different Distribution Mixed Model and Its Parameters

Zhang Bo1, Liu Hefei2, Wang Kun2   

  1. 1.Party Committee Graduate Work Department,Yunnan University, Kunming Yunnan 650504,China;
    2.School of Mathematics and Statistics,Qujing Normal University, Qujing Yunnan 655011, China
  • Received:2018-09-11 Published:2018-11-26

摘要: 对常见的混合模型进行了推广,提出了异分布混合模型的概念.以二项分布和泊松分布为例,详细给出了异分布混合模型的数学定义,以及异分布混合模型参数的贝叶斯估计.在参数估计中,选择狄尼克莱分布、贝塔分布和伽玛分布作为相关参数的共轭先验分布,利用MCMC算法对模型的参数进行了后验模拟,用后验均值作为参数的估计值.最后将参数的估计结果与模型的真实值进行比较,证明了估计结果的可靠性.

关键词: 异分布混合, 贝叶斯估计, 先验分布, MCMC算法

Abstract: We generalize the common hybrid models and put forward the concept of heterogeneous mixed models. Taking the binomial and Poisson distributions as an example, the mathematical definitions of the heterogeneous mixed models are given in detail. In the parameter estimation, we choose the Dignley distribution, the beta distribution and the gamma distribution as the conjugate prior distribution of the relevant parameters, and use the MCMC algorithm to carry on the posteriori simulation of the model parameters, using the posterior mean value as the parameter Finally, the result of parameter estimation is compared with the real value of the model, which proves the reliability of the estimation result.

Key words: Different distribution Mixed model, Bayesian estimation, Prior distribution, MCMC algorithm

中图分类号: