部分线性变系数Tobit模型的稳健贝叶斯统计推断

Robust Bayesian Statistical Inference of Partially Linear Varying Coefficient Tobit Model

提出了一种基于重尾分布假定的部分线性变系数Tobit回归模型,并结合贝叶斯理论和MCMC算法给出了模型参数的统计推断结果。最后,数值模拟和实证分析表明,与正态分布假设下的部分线性变系数Tobit模型相比较,所提出的统计模型能够得到更稳健、更有效的统计推断结果。

This paper presents a partially linear varying coefficient Tobit regression model based on heavy-tailed distribution assumption,and then gives the statistical inference results of model parameters by combining Bayesian theory and MCMC algorithm.Finally,the results of numerical simulation and empirical analysis show that the pro-posed method can obtain more robust and effective statistical inference than that of the partial linear varying coeffi-cient Tobit model under the assumption of normal distribution.