摘要
针对两步法在多个结构突变估计中存在的过度估计缺陷,文章通过BIC准则确定分段线性Tobit回归模型中的突变个数,利用线性化技术解决含突变时目标函数不可导的困难,得到了突变个数、突变位置及模型参数估计。模拟结果表明新方法具备有效性和稳健性,用该方法进行的实证结果显示,家庭金融资产会受到受教育水平的正向影响,存在两个结构突变,受教育水平在高中前后对家庭金融资产的提升作用较为明显,在本科前后对家庭金融资产的提升作用有所降低。
Aiming at the overestimation defect of the two-step method in the estimation of multiple structural mutations, this paper determines the number of mutations in the piecewise linear Tobit regression model by BIC criterion, and then uses linearization technique to solve the difficulty that the objective function cannot be differentiated with mutations, obtaining the number of mutations, the location of mutations and the parameter estimation of the model. The simulation results show that the new method is effective and robust. The empirical results of this method show that the family financial assets are positively affected by the level of education, with two structural mutations, that the level of education has an obvious promotion effect on the family financial assets before and after high school, and that it has a lower promotion effect on the family financial assets before and after college.
作者
王小刚
李冰
Wang Xiaogang;Li Bing(School of Mathematics and Information Sciences,North Minzu University,Yinrhuan 750021.China)
出处
《统计与决策》
CSSCI
北大核心
2021年第19期21-25,共5页
Statistics & Decision
基金
国家社会科学基金资助项目(17BTJ033)。
