摘要
Logistic回归模型是一种广义线性模型,广泛应用于二分类数据的建模问题之中,经典的Logistic回归模型通常基于二项分布进行建模,在某些实际问题中,二项分布是不合适的.在平方损失函数下,研究了具有Logistic回归结构的几何分布参数的贝叶斯估计问题,选取正态分布作为参数的先验分布,运用MCMC方法在WinBUGS软件中进行gibbs抽样,得到参数的后验样本.数值模拟表明估计效果良好,最后将所提出的模型应用于米其林餐厅定级数据当中.
The Logistic regression model,one of the generalized linear models,is widely used in modeling the binary data set.The classical Logistic regression models are usually based on the binomial distribution.In some specific problems,binomial distribution is not appropriate.In this paper,we considered the Bayesian estimation problem for a geometric distribution with Logistic regressive structure under the quadratic loss.The standard normal priors are chosen and MCMC method is used for estimating the model parameters.Gibbs sampling method is used for drawing the posterior sample under the WinBUGS software.Simulation results show that the estimation performs well.Finally,the proposed model is applied to the Michelin restaurant grading data set.
出处
《辽宁师范大学学报（自然科学版）》
CAS
2016年第3期295-298,共4页
Journal of Liaoning Normal University:Natural Science Edition
基金
国家自然科学基金资助项目(11271155)
吉林省自然科学基金资助项目(20130101066JC
20101596)
