半参数模型的设定与应用研究

发布时间：2018-01-20 02:40

本文关键词： 模型设定非参数模型半参数模型残差回归检验核估计技术　出处：《天津财经大学》2016年博士论文　论文类型：学位论文

【摘要】：模型设定是计量经济学建模过程的重要组成部分。计量经济学的本质就是用数据来描述经济现象,最终体现经济数据的变化规律并用模型形式来表达。在计量经济学模型完整的研究框架中,模型设定居于建模过程的首位。当模型被正确地设定时,基于该模型得到的估计和检验结果才能应用到经济数据分析中。只有计量经济模型真实地反映经济系统运行的规律,才能为经济政策评价和经济系统结构分析提供可靠的基础。一旦模型被错误设定,那么任何的估计和检验方法都无法弥补这种模型结构上的错误。此时,由模型得到的结论就不能如实地刻画经济数据的变化特征,并且很有可能是错误的,这会给经济研究带来很大的风险。因此,对模型的设定问题进行研究是非常必要的,且具有重要的理论意义与应用价值。模型设定是指在建立计量经济模型时,对变量选择、回归函数形式以及误差项分布等方面进行地讨论。这里主要研究的问题是对回归函数形式的确定。半参数回归模型是一类重要的计量经济模型,不仅兼顾了参数回归模型和非参数回归模型的优点,同时还能降低参数回归模型对数据假设条件的依赖程度。由于半参数回归模型的结构能有效地缓解非参数回归模型的维数陷阱问题,使得半参数回归模型可广泛地应用于社会生活的各个领域中。因此,从理论和应用的角度来看,半参数回归模型的研究都具有重要的意义。同参数回归模型类似,半参数回归模型也存在模型被错误设定的可能。由于半参数回归模型的理论还不够完善,且有广泛的应用前景,因此文章将半参数回归模型的设定作为重点研究问题。文章首先对国内外非参数回归模型与半参数回归模型设定方法的相关文献进行了系统地梳理和归纳,充分了解了模型设定在国内外的研究现状,并重点研究了半参数回归模型的模型设定内容,为文章的选题提供了理论支撑。基于Yatchew(2003)提出的非参数回归模型设定研究的基本框架,将半参数回归模型中的部分线性模型和单指标模型的模型设定问题也纳入到该研究框架内,事实上,对这些问题的研究最终都要归纳到对残差的回归分析上。这项工作扩展了模型设定的研究范围,为半参数回归模型的设定方法问题提供了一个新的研究方法。然后文章通过深入地研究简化了各种设定方法的步骤,为使用它们的研究者提供了很好的工具。在实证方面,将文章的研究成果应用到我国上市公司主要财务指标与股票价格的研究中。通过模型的设定方法得到了适合分析上市公司主要财务指标与股票价格之间关系的计量模型,为这方面的研究提供了理论依据。文章的创新点主要体现在以下三点:一、对国内外非参数模型和半参数模型的设定方法相关文献进行了系统地梳理与归纳。按研究方法的不同,对各种方法的适应性进行了探讨,使得非参数回归模型和半参数回归模型的设定理论更加完善,为应用这些模型的研究者提供了简便的方法。二、对半参数回归模型中的部分线性模型和单指标模型的设定方法进行了改进,将这两类模型的检验过程纳入了一般的研究框架中,进一步完善了半参数回归模型的设定方法。三、将文章的研究成果应用于我国上市公司主要财务指标与股票价格的实际研究中,利用半参数模型对它们进行分析,得到了较好的结论。在此基础上,与实证分析中常采用的参数线性模型进行了比较和研究,得到了半参数模型比参数模型以及非参数模型效果更好的结论。此项研究不仅对我国上市公司财务指标与股票价格的模型建立提供了理论基础,还为半参数模型的设定问题提供了很好的实证依据。
[Abstract]:The model set is an important part of econometric modeling process. The essence of econometrics is to use data to describe economic phenomena, and ultimately reflect the variation of economic data and the model form of expression. In the research framework of the econometric model in the complete model. The modeling process is settled in the first place. When the model is correctly set based on this model, the estimation and analysis of test results can be applied to economic data. Only the econometric model to reflect the real economic system operation rules, in order to economic policy and economic evaluation system structure analysis to provide reliable basis. Once the model is wrong setting, then estimate and test methods all can compensate for this the model structure error. At this time, the conclusions can not truthfully describe the change characteristics of economic data, and it is likely Is wrong, it will bring great risk to economic research. Therefore, it is very necessary to study the problem of setting model, and has important theoretical significance and practical value. The model set is defined in an econometric model, to select variables for the regression function type and error distribution etc. the discussion here. The main issue is to determine the form of the regression function. The semi parametric regression model is a class of econometric models is important, not only taking into account the parametric regression model and nonparametric regression model of the advantages, but also reduced the dependence of parameters on the assumption of data regression models. Due to the structure of the semi parametric regression model can effectively alleviate the dimension trap problem of non parametric regression model, each field makes the semiparametric regression model can be widely used in social life. Therefore, from the theory and The application's point of view, is very important to study the semiparametric regression model. The regression model is similar, there are set of possible error model of semi parametric regression model. Because of the semiparametric regression model theory is still not perfect, and has a wide application prospect, so the semiparametric regression model as set focus on the problem. A systematic combing and summarizing related literature at home and abroad based on the nonparametric regression model and semiparametric regression model setting method, fully understand the present research situation at home and abroad set model, set the content and focus on the semiparametric regression model, and provides theoretical support for the the topic is based on Yatchew. (2003) the basic framework set of nonparametric regression model, semi parametric regression model in partial linear model and single index model The model set is also incorporated into the research framework, in fact, the research on these problems will eventually be summed up to the regression analysis of the residuals. This work expands the research scope of the model set, provides a new research method for the setting method of semiparametric regression model. Then through in-depth study on simplified method of setting various steps, provides a good tool for the research of using them. In the empirical analysis, the article will study the application of research results to the listed companies in China the main financial indicators and stock price. The econometric model suitable for the analysis of listed companies between the main financial indicators and stock prices are through the method of setting model, provides a theoretical basis for the research. The innovation of this paper is mainly reflected in the following three points: first, the domestic and foreign non parametric model and semi parametric The systematic induction method of setting model related literature. According to the different research methods, the adaptability of various methods are discussed, the set theory makes the nonparametric regression model and semiparametric regression model is more perfect, the researchers used these models provide a simple method. The two part, setting method of linear model and single index model of semi parametric regression was improved, the inspection process of the two kinds of models into a general research framework, to further improve the setting method of semiparametric regression model. Three, the actual research results application of the listed companies in China's main financial index and stock price in using semi parametric model to analyze them, and get better conclusions. On this basis, compared with the linear model parameters used in empirical analysis And research, obtained the semi parametric model than the parametric model and non parametric model better conclusion effect. This research not only on the financial index and stock price of Listed Companies in China the model provides a theoretical basis and provides empirical evidence for the good set problem for semi parametric model.

【学位授予单位】：天津财经大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：F224;F832.51;F275

【相似文献】