随机效应半参数二值响应模型的估计研究

发布时间：2018-05-17 09:49

本文选题：随机效应半参数二值响应模型 + B样条估计　；参考：《管理科学学报》2015年11期

【摘要】：为了研究诸如在收入差距和健康关系中某些解释变量的影响效应可能依赖于其他解释变量的状况,降低可能存在的模型设定和遗漏变量偏误,本文提出了随机效应半参数二值响应模型,其中非参数的设定还可用于数据的初探性分析.随后本文探讨了模型非参数和参数部分的估计.由于似然函数中随机效应和非参数的存在加大了估计难度,为此本文采用B样条方法逼近非参数部分,同时将随机效应视为缺失数据,利用EM算法和MCMC方法建立了模型参数的估计,并研究了其一致性.模拟研究结果表明估计量在有限样本下的表现良好,最后将模型运用于收入差距与健康关系的实例研究中,结果表明数据支持收入差距弱假说.
[Abstract]:In order to study the effect of certain explanatory variables, such as in the income gap and health relationship, which may depend on the condition of other explanatory variables, and to reduce the possible errors of model setting and missing variables, In this paper, a semi-parametric binary response model with random effect is proposed, in which non-parametric setting can also be used for the preliminary analysis of data. Then this paper discusses the estimation of nonparametric and parametric parts of the model. Because the existence of random effect and nonparametric in likelihood function makes it more difficult to estimate, the B-spline method is used to approximate the nonparametric part, and the random effect is regarded as missing data. The estimation of model parameters is established by using EM algorithm and MCMC method, and its consistency is studied. The simulation results show that the estimator performs well in a limited sample. Finally, the model is applied to the case study of the relationship between income gap and health. The results show that the data support the weak hypothesis of income gap.
【作者单位】：上海财经大学经济学院;上海财经大学数理经济学重点实验室;
【基金】：教育部“新世纪优秀人才支持计划”资助项目(NECT-10-0562) 国家自然基金资助项目(11271242) 上海财经大学“讲席教授”资助项目
【分类号】：F124.7;F224

【参考文献】