半参数面板回归模型的变点分析

发布时间：2018-01-13 14:30

本文关键词：半参数面板回归模型的变点分析　出处：《大连理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：在这篇文章中,我们研究的模型是存在变点的面板数据的半参数模型,其中变点发生在某个未知的时间或未知位置。半参数模型和面板数据是经济计量学的两个热点问题,将半参数模型与面板数据结合起来具有很重要的研究意义,此外考虑到实际问题中可能发生变化的情况,本文又在此基础上引入了变点,使得模型能够应用到更广的领域。首先,我们将面板数据的半参数模型和有变点的线性模型结合,提出了有变点的面板数据的半参数模型,对于单变点的情况,我们提出了两种不同形式的模型：一种是变点只发生在线性部分；另一种情况是变点发生在每一个半参数部分。我们的首要任务是将非参数部分转化为参数形式的表达式,这里我们分别采用两阶段局部估计法和多项式样条法来解决这个问题,然后要给出变点的估计,这里采用常见的最小二乘方法来估计变点。其次,我们分别做了单变点和两变点的模拟,并将其与线性模型的变点估计情况进行比较。从模拟的结果中我们发现,与有变点的面板数据线性模型相比较,随着样本容量的增大,面板数据的半参数模型的变点位置趋于稳定的速度更快,并且回归系数的估计也更加准确。从模拟结果可以得出本文模型的建立是有意义的。最后,我们进行了实证分析。主要运用中国12家银行在2008年至2014年间,股东回报率、资本充足率以及流动性比率的面板数据,通过对三者的关系进行分析,我们给出变点的估计值和系数的估计值。
[Abstract]:In this paper, we study a semi-parametric model of panel data with variable points. The semi-parametric model and panel data are two hot topics in econometrics. It is very important to combine semi-parametric model with panel data. In addition, considering the possible changes in the actual problems, this paper introduces the change points on this basis, which makes the model can be applied to a wider range of fields. First of all. We combine the semi-parametric model of panel data with the linear model with variation points, and propose a semi-parametric model of panel data with variable points, for the case of single change point. We propose two different models: one is that the change points occur only in the linear part; In another case, the change occurs in every semi-parametric part. Our first task is to convert the nonparametric part into an expression in the form of a parameter. Here we use the two-stage local estimation method and polynomial spline method to solve this problem, then we give the estimation of the change point, here we use the common least square method to estimate the change point. Secondly, we use the least square method to estimate the change point. Second, we use the least square method to estimate the change point. We do the simulation of single change point and two change point respectively, and compare them with the change point estimation of linear model. From the simulation results, we find that it is compared with the panel data linear model with the change point. With the increase of sample size, the change point position of the semi-parametric model of panel data tends to stabilize faster. And the estimation of regression coefficient is more accurate. From the simulation results can be concluded that the establishment of this model is meaningful. Finally. We have conducted empirical analysis. We mainly use the panel data of 12 Chinese banks from 2008 to 2014 on the return of shareholders, capital adequacy ratio and liquidity ratio. Through the analysis of the three relations, we give the estimated values of the change points and the coefficients.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F832.3;F224

【共引文献】