分位数因子增广的分位数自回归分布滞后模型构建  

Construction of Quantile Factor-augmented Quantile Autoregression Distributed Lag Model

作　　者：黄玉婷 傅德印 Huang Yuting;Fu Deyin(School of Statistics and Data Science,Lanzhou University of Finance and Economics,Lanzhou 730020,China;School of Labor Economics,China University of Labor Relations,Beijing 100048,China)

机构地区：[1]兰州财经大学统计与数据科学学院,兰州730020 [2]中国劳动关系学院劳动经济学院,北京100048

出　　处：《统计与决策》2024年第12期35-41,共7页Statistics & Decision

基　　金：甘肃省优秀博士生项目(23JRRA1189);甘肃省研究生“创新之星”项目(2023CXZX-700);兰州财经大学博士研究生科研创新项目(2022D02,2022D05)。

摘　　要：因子增广回归是利用高维数据对宏观经济进行预测的重要方法。然而,均值回归框架下的因子模型和回归模型在面对异常值或厚尾分布时结果不够稳健。为此,文章在分位数回归框架下构建分位数因子增广分位数自回归分布滞后模型,该模型通过构建分位数因子模型对高维数据进行降维,提取不同分位点的公共因子;进一步,利用提取出的公共因子和响应变量的滞后项构建分位数自回归分布滞后模型。数值模拟结果表明,在数据出现异常值或厚尾分布的情境下,分位数因子分析估计的均值因子和非均值因子更稳健,分位数因子增广回归的预测性能优于因子增广回归的预测性能,且分位数因子增广分位数自回归分布滞后模型的预测性能优于基准模型。Factor-augmented regression is an important method to forecast macroeconomics using high dimensional data.However,the results of factor models and regression models under the mean regression framework are not robust enough in the face of outliers or thick tail distributions.In view of this,the paper constructs a quantile factor-augmented quantile autoregression distributed lag model under quantile regression frame.In this model,the quantile factor model is constructed to reduce the dimensionality of high-dimensional data,with the common factors of different quantiles extracted.Further,the quantile autoregressive distributed lag model is constructed by using the extracted common factors and lag terms of response variables.Numerical simulation results show that the mean and non-mean factors estimated by quantile factor analysis are more robust in the case of data outliers or thick tail distribution.The predictive performance of quantile factor-augmented regression is better than that of factor-augmented regression,and the predictive performance of quantile factor-augmented autoregressive distributed lag model is better than that of the benchmark model.

关 键 词：分位数因子 分位数回归 因子增广回归 自回归分布滞后模型 

分 类 号：F224[经济管理—国民经济]