高维面板数据中因子个数选择方法及其模拟研究

发布时间：2018-05-07 21:57

本文选题：面板数据 + 因子模型　；参考：《浙江工商大学》2014年硕士论文

【摘要】：因子模型由于能够简化高维数据的分析过程,从而广泛的应用于经济、金融、生态、医药等领域。而准确的选择因子个数始终是因子模型应用上的核心问题。近些年来,大量关于因子个数确定的文献频繁见诸统计、计量等学术期刊,本文通过Monte Carlo比较研究三种经典的面板因子个数选择准则的有限样本性质,并基于其中两种邻近特征值比形式的准则,构造了两种新的因子个数选择准则,Monte Carlo模拟表明,新的准则在部分弱因子情形下有更好的稳健性和适应性。考虑到以上准则假定因子个数不随着样本量的改变而变化,因此本文发展了上述的信息准则以适应可变因子数目下的数据情形。最后,以沪深300指数成分股中的226支连续交易288天的股票日收益率序列构建平衡面板数据集,使用不同因子个数选择方法,得出存在两个因子的结论,并发现沪深300指数收益序列和因子高度相关,进一步地以本文发展的适用于可变因子数目的准则研究发现,所考察区间内的样本数据始终由两个因子驱动,不存在经济结构变动。
[Abstract]:Because the factor model can simplify the analysis process of high dimensional data, it is widely used in the fields of economy, finance, ecology, medicine and so on. Accurate selection of the number of factors is always the core problem in the application of the factor model. In recent years, a large number of literatures on factor number determination have been published frequently in academic journals, such as statistics and econometrics. Through Monte Carlo, this paper compares the finite sample properties of three classical criteria for selecting the number of panel factors. Based on two adjacent eigenvalue ratio criteria, two new factor selection criteria are constructed. The results show that the new criterion has better robustness and adaptability in the case of partial weak factors. Considering that the above criterion assumes that the number of factors does not change with the change of sample size, the above information criterion is developed to adapt to the data situation under the variable number of factors. Finally, the data set of balance panel is constructed by the daily yield sequence of 226 stocks traded in Shanghai and Shenzhen 300 index for 288 days, and the conclusion that there are two factors is obtained by using the method of selecting different number of factors. It is also found that the return sequence of CSI 300 index is highly correlated with factors. Further, based on the criteria developed in this paper for the number of variable factors, it is found that the sample data in the investigated interval are always driven by two factors. There is no change in economic structure.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51;F224

【参考文献】