基于STAR模型单位根检验的研究及实证分析

发布时间：2018-05-03 02:18

本文选题：单位根检验 + 经验似然比统计量　；参考：《浙江工商大学》2017年硕士论文

【摘要】：单位根检验是计量经济学中十分重要的研究内容之一。尤其在实际的金融时间序列多为非线性,并且大多数是含有单位根的非平稳序列的背景下,平稳性检验在研究时间序列数据中已是必不可少的一个步骤。但是传统的单位根检验是基于线性模型提出的,针对现今众多非线性模型的检验效果并不是十分有效,容易造成过分接受非平稳的假设,从而引起误判。基于此,本文针对实际应用较多的非线性STAR模型进行了相应的单位根检验研究。考虑到STAR模型在实际拟合时间序列数据时,模型的残差项常服从GARCH过程,因此本文在前人的基础上构建了检验似然比检验统计量l(δ)。似然比检验统计量极大的提高了 STAR模型的单位根检验功效,并且与汪卢俊(2014)提出的针对LSTAR-GARCH模型的单位根检验统计量tNG相比,避免了计算估计方差,有效的降低了计算复杂度,提高了估计统计量的稳定性。本文首先是对单位根检验的历史和理论进行了介绍,然后基于汪卢俊(2014)提出的针对LSTAR-GARCH模型的单位根检验统计量tNG,在文中第三章创造性的提出了基于LSTAR-GARCH模型的经验似然比检验统计量l(δ),并推导出其极限分布。其中关于时间序列的条件方差时变性特征(GARCH项),tNG的极限分布在推导过程中需要考虑到tNG的估计方差,这样会增加tNG的不稳定性和计算复杂度,而经验似然比检验统计量可以有效地避免计算统计量的估计方差,从而提高单位根检验的效果。为了验证第三章中的理论,本文第四章通过蒙特卡罗和Bootstrap方法进行模拟和功效比较,在模拟的角度进一步的说明这一情况。更进一步,第五章结合我国上证指数股票数据进行实证分析,通过拟合情况来比较,说明使用经验似然比检验统计量检验,构建的STAR模型最为准确,能够为投资者提供更可靠的信息。
[Abstract]:Unit root test is one of the most important research contents in econometrics. Especially under the background that the actual financial time series are mostly nonlinear and most of them are non-stationary sequences with unit roots, the stationary test is an essential step in the study of time series data. However, the traditional unit root test is based on the linear model. The test effect for many nonlinear models is not very effective. It is easy to overaccept the assumption of non-stationary, thus causing misjudgment. Based on this, this paper studies the unit root test of nonlinear STAR model which is widely used in practice. Considering that the residual term of the STAR model is usually followed by the GARCH process when the time series data are fitted, the test likelihood ratio test statistic L (未) is constructed on the basis of previous studies. Likelihood ratio test statistics greatly improve the efficiency of unit root test of STAR model, and compared with the unit root test statistic tNG for LSTAR-GARCH model proposed by Wang Lujun 2014, it avoids the estimated variance and reduces the computational complexity effectively. The stability of estimation statistics is improved. This paper first introduces the history and theory of unit root test. Then, based on the unit root test statistic for LSTAR-GARCH model proposed by Wang Lujun (2014), the empirical likelihood ratio test statistic based on LSTAR-GARCH model is creatively proposed in chapter 3, and its limit distribution is deduced. For the conditional variance of time series, it is necessary to take into account the estimated variance of tNG in the derivation of the limit distribution of the term GARCH, which will increase the instability and computational complexity of tNG. The empirical likelihood ratio test statistics can effectively avoid calculating the estimated variance of the statistics and thus improve the effect of unit root test. In order to verify the theory in the third chapter, the fourth chapter of this paper uses Monte Carlo and Bootstrap methods to simulate and compare the effectiveness of the simulation, in order to further explain this situation in the perspective of simulation. Furthermore, the fifth chapter combines the stock data of Shanghai Stock Exchange of China to carry on the empirical analysis, through the comparison of the fitting situation, shows that using the empirical likelihood ratio test statistic test, the STAR model is the most accurate. To provide investors with more reliable information.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F224;F832.51

【参考文献】