摘要
样本数据中经常会出现大量取值为零的观测值,导致数据过度离散.为解决传统模型不能很好地进行预测与拟合的问题,提出了零膨胀几何分布回归模型,建立了响应变量与协变量之间的关系,通过巧妙地引入隐变量并进行数据扩充,并基于EM算法下的极大似然估计法和贝叶斯方法对回归参数向量进行估计,利用数值模拟的方法对两种估计方法的成效进行了评价.
There are usually many zero observations in the sample data,which may lead to excessive dispersion of data.In order to solve the problem that the traditional model can not forecast and fit well,this paper proposes a zero-inflated geometric distribution regression model.It establishes the relationship between response variables and covariates and tactfully introduces latent variables and uses data augmentation strategy.The study also estimates the regression parameter vectors through the maximum likelihood estimation based on EM algorithm and Bayesian estimation respectively and compares the performance of the two estimation methods by numerical simulation.
作者
肖翔
王国强
刘福窑
刘莹泽
Xiao Xiang;Wang Guoqiang;Liu Fuyao;Liu Yingze(School of Mathematics,Physics and Statistics,Shanghai University of Engineering Science,Shanghai 201620)
出处
《嘉兴学院学报》
2020年第6期25-30,共6页
Journal of Jiaxing University
基金
国家自然科学基金面上项目(11971302)
上海市高等教育学会规划项目(GJEL1817)
上海工程技术大学教育科学研究项目(y201821001)
上海工程技术大学大学生创新训练项目(CX1921002)。
关键词
零膨胀几何分布
回归模型
数据扩充
EM算法
贝叶斯估计
zero-inflated geometric distribution
regression model
data augmentation
EM algorithm
Bayesian estimation
