宏观经济序列的混频特征为建模和预测带来了困难和挑战, 而且数据中天然存在的某些组结构决定了组特殊因子的存在, 为此本文在已知组结构下研究了混频时间序列的因子分析方法.本文首先在同频时间序列下给出组结构因子模型, 提出基于LF方法的两步估计.之后在混频时间序列下借鉴EM算法的框架, 提出估计组结构因子模型的MFGF-LF方法, 本质是将因子模型和缺失数据看作是相互依存的结构, 以迭代估计的方式估计因子并且补齐缺失值.模拟数据和实际数据的分析显示, 相比于MFGF-PCA、EM-LF和EM-PCA方法, 本文提出的MFGF-LF方法具有较低的估计和预测误差, 说明考虑组结构的因子分析明显优于普通的因子分析, 而且在因子提取方面LF方法比PCA方法更有优势.
Abstract
Mixed frequency in the macroeconomic time series brings difficulties and challenges to modeling and forecasting, and grouped structures in the data determine the existence of group-specific factors. For these two reasons, we studied the factor analysis of mixed frequency time series under the known group structure. Firstly, the grouped factor model is introduced under completely observed time series, and a two-step estimation based on the LF method is proposed. Then, MFGF-LF method for estimating the grouped factor model in the mixed frequency time series is proposed based on the EM algorithm. The essence of this method is to treat the factor model and the missing data as interdependent structure, the iterative method will estimates the factor and also imputes the missing data. The analysis of simulated data and actual data shows that, compared with the MFGF-PCA, EM-LF and EM-PCA methods, the MFGF-LF method proposed in this paper has lowest estimation error and prediction error, indicating that the factor analysis with the group structure is obviously better than ordinary factor analysis and the LF method has more advantages than PCA method in terms of factor extraction.
关键词
组结构 /
混频序列 /
因子模型
{{custom_keyword}} /
Key words
grouped structure /
mixed frequency time series /
factor model
{{custom_keyword}} /
中图分类号:
F224
{{custom_clc.code}}
({{custom_clc.text}})
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Stock J H, Watson M W. Macroeconomic forecasting using diffusion indexes[J]. Journal of Business & Economic Statistics, 2002, 20(2): 147-162.
[2] Sweeney J C, Soutar G N. Consumer perceived value: The development of a multiple item scale[J]. Journal of Retailing, 2001, 77(2): 203-220. [1] Fama E F, French K R. Common risk factors in the returns on stocks and bonds[J]. Journal of Financial Economics, 1993, 33: 3-56.
[3] 倪宣明, 郑田田, 孙会霞, 等. 统计因子、 经济因子, 殊途同归?[J]. 系统工程理论与实践, 2023, 43(12): 3385-3406. Ni X M, Zheng T T, Sun H X, et al. Can statistical factors and economic factors reach the same goal in different ways?[J]. Systems Engineering—Theory & Practice, 2023, 43(12): 3385-3406.
[4] 姜富伟, 薛浩, 周明. 大数据提升了多因子模型定价能力吗?——基于机器学习方法对我国A股市场的探究[J]. 系统工程理论与实践, 2022, 42(8): 2037-2048. Jiang F W, Xue H, Zhou M. Does big data improve multi-factor asset pricing models? Exploration of China's A-share market with machine learning[J]. Systems Engineering—Theory & Practice, 2022, 42(8): 2037-2048.
[5] 冯玉林, 汤珂, 康文津. 中国大宗商品期货市场定价机制研究[J]. 金融研究, 2022(12): 149-167. Feng Y L, Tang K, Kang W J. The pricing mechanism in Chinese commodity futures markets[J]. Journal of Financial Research, 2022(12): 149-167.
[6] Bai J, Ng S. Determining the number of factors in approximate factor models[J]. Econometrica, 2002, 70(1): 191-221.
[7] Forni M, Hallin M, Lippi M, et al. The generalized dynamic factor model: One-sided estimation and forecasting[J]. Journal of the American Statistical Association, 2005, 100(471): 830-840.
[8] Lam C, Yao Q, Bathia N. Estimation of latent factors for high-dimensional time series[J]. Biometrika, 2011, 98(4): 901-918.
[9] Chen P. A grouped factor model[R]. Munich: University Library of Munich, 2010.
[10] Chen P. Common factors and specific factors[R]. Munich: University Library of Munich, 2012.
[11] Bai J, Ando T. Multifactor asset pricing with a large number of observable risk factors and unobservable common and group-specific factors[R]. Munich: University Library of Munich, 2013.
[12] Schumacher C, Breitung J. Real-time forecasting of German GDP based on a large factor model with monthly and quarterly data[J]. International Journal of Forecasting, 2008, 24(3): 386-398.
[13] 秦磊, 郁静, 孙强. 混频时间序列的潜在因子分析及其应用[J]. 统计研究, 2019, 36(9): 104-114. Qin L, Yu J, Sun Q. Latent factor analysis of mixing time series and application[J]. Statistical Research, 2019, 36(9): 104-114.
[14] Giannone D, Reichlin L, Small D. Nowcasting: The real-time informational content of macroeconomic data[J]. Journal of Monetary Economics, 2008, 55(4): 665-676.
[15] Bai J, Ng S. Matrix completion, counterfactuals, and factor analysis of missing data[J]. arXiv Preprint arXiv: 1910.06677, 2019.
[16] Jin S, Miao K, Su L. On factor models with random missing: EM estimation, inference, and cross validation[J]. Journal of Econometrics, 2021, 222(1): 745-777.
[17] Andreou E, Gagliardini P, Ghysels E, et al. Inference in group factor models with an application to mixed-frequency data[J]. Econometrica, 2019, 87(4): 1267-1305.
[18] Boivin J, Ng S. Are more data always better for factor analysis?[J]. Journal of Econometrics, 2006, 132(1): 169-194.
[19] Tibshirani R, Saunders M, Rosset S, et al. Sparsity and smoothness via the fused lasso[J]. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2005, 67(1): 91-108.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
国家社会科学基金重大项目(22&ZD072);对外经济贸易大学研究生科研创新基金(202475)
{{custom_fund}}