NMAR机制下非线性均值方差模型的Bayes估计  

Bayesian Estimation for Nonlinear Mean and Variance Models with Nonignorable Missing Data Mechanism

作　　者：赵远英[1] 唐安民 段星德 庞一成[3] ZHAO Yuan-ying;TANG An-min;DU AN Xing-de;PANG Yi-cheng(College of Mathematics and Information Science,Guiyang University,Guiyang 550005,China;Department of Statistics,Yunnan University,Kunming 650091,China;School of Mathematics and Statistics,Guizhou University of Finance and Economics,Guiyang 550025,China)

机构地区：[1]贵阳学院数学与信息科学学院,贵州贵阳550005 [2]云南大学统计系,云南昆明650091 [3]贵州财经大学数学与统计学院,贵州贵阳550025

出　　处：《数学的实践与认识》2019年第21期157-165,共9页Mathematics in Practice and Theory

基　　金：国家自然科学基金(11761016,11501073,11561074,11661015);贵阳市科技局贵阳学院科技专项基金(GYU-KYZ(2019-2020)PT04-04)

摘　　要：讨论响应变量带有不可忽略缺失数据的非线性均值方差模型的Bayes估计问题.缺失数据机制由logistic回归模型来指定,运用Gibbs抽样及MH算法得到模型参数和缺失数据机制参数的联合Bayes估计,模拟研究和实例分析展示上述模型和方法的可行性.A Bayesian approach is proposed to analyze nonlinear mean and variance models,in which the responses are missing with non-ignorable missingness mechanism.The missingness mechanism is specified by a logistic model,Gibbs sampler and Metropolis-Hastings algorithm is employed to obtain the joint Bayesian estimates of model parameters and missingness mechanism parameters.The proposed methodology is demonstrated by several simulation studies and a real example.

关 键 词：BAYES估计 GIBBS抽样 MH算法 不可忽略缺失数据机制 非线性均值方差模型 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]