大维因子模型的因子个数估计

发布时间：2018-03-17 02:22

本文选题：广义动态因子模型　切入点：渐近静态因子模型　出处：《浙江工商大学》2017年硕士论文　论文类型：学位论文

【摘要】：本文分别对大维条件下的广义动态因子模型和渐近静态因子模型构造了一种信息准则以估计他们的因子个数。本文主要受Onatski(2009)的启发,其核心在于以下两点:第一,关于广义动态因子模型的方法,在一定的条件下,利用傅立叶变换将序列的动态结构转化为类似Chamberlain和Rothschild(1983)提出的静态结构,然后用类似Bai和Ng(2002)的方法建立信息准则,相比Hallin和Liska(2007)的方法,我们的条件更加宽松。第二,关于渐近静态因子模型的方法,我们对原变量做一个特殊的变换:为变换后的向量,j=1,…,T/2),然后在一定条件下建立信息准则。我们的方法的一致性在文中的假设下均得到了有效的保证,而且蒙特卡洛模拟显示,我们的方法比Bai和Ng(2002)和Haliin(2007)更能适应强相关和强噪声的情况。本文的主要内容如下:第一章主要论述因子模型的研究背景、意义和研究现状,并提出广义动态因子模型和渐近静态因子模型;这两个模型的因子个数估计方法及假设和证明分别在第二章和第三章;第四章利用模特卡罗模拟拿我们的方法与其他方法进行比较;第五章是我们的结论。
[Abstract]:In this paper, we construct an information criterion for generalized dynamic factor model and asymptotically static factor model under the condition of large dimension to estimate the number of their factors. This paper is mainly inspired by Onatskii 2009), the core of which lies in the following two points: first, On the method of generalized dynamic factor model, under certain conditions, the dynamic structure of a sequence is transformed into a static structure similar to Chamberlain and Rothschild1983by Fourier transform, and then the information criterion is established by using a method similar to Bai and Ngn 2002). Our conditions are more relaxed than those of Hallin and Liska 2007. Second, the method of asymptotic static factor model, We make a special transformation of the original variable: we set up the information criterion under certain conditions for the transformed vector / t / 2. The consistency of our method is effectively guaranteed under the assumptions in this paper, and the Monte Carlo simulation shows that, Our method is more suitable for strong correlation and strong noise than Bai and Nggan 2002) and Haliin 2007. The main contents of this paper are as follows: the first chapter mainly discusses the background, significance and research status of the factor model. Then the generalized dynamic factor model and the asymptotic static factor model are proposed, and the methods of estimating the number of factors and their hypotheses and proofs are given in chapter 2 and chapter 3, respectively. Chapter 4th uses model simulations to compare our methods with other methods; Chapter 5th is our conclusion.
【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

【相似文献】