1.School of Mathematics and Statistics, Chongqing Technology and Business University, Chongqing 400067, China
2.Chongqing Key Laboratory of Statistical Intelligent Computing and Monitoring, Chongqing 400067, China
Show less
文章历史+
Received
Accepted
Published
2023-09-13
2024-03-07
2026-09-25
Issue Date
2026-09-16
PDF (722K)
摘要
本文基于弹性网(Elastic net)惩罚为面板数据固定效应线性回归模型提出了一种新的变量选择方法,该方法能处理强相关变量的变量选择问题。先通过正向正交偏差变换消除固定效应,然后构造了基于Elastic net惩罚的最小二乘目标函数,进而同时进行回归系数的估计和变量选择,证明了所得估计的组效应性质。所提出的变量选择过程不仅消除了固定效应的影响,而且即使协变量存在共线性时,也能很好地识别出真实模型。其次,模拟研究了本文提出方法的有限样本性质。模拟研究表明:与最小绝对收缩和选择算子(Least Absolute Shrinkage and Selection Operator,Lasso),自适应最小绝对收缩和选择算子(Adaptive Least Absolute Shrinkage and Selection Operator,ALasso)以及平滑截断绝对偏差惩罚(Smoothly Clipped Absolute Deviation, SCAD)方法相比,提出的Elastic net变量选择方法能够很好地将强相关变量选入模型。最后,采用文中提出的方法分析了实际数据。
Abstract
Based on Elastic net, a new variable selection method is proposed in this paper for linear panel data models with fixed effects. The variable selection problem with the strongly correlated variables can be dealt by the proposed variable selection method. Firstly, the fixed effect is eliminated by the forward orthogonal deviation transformation. Then, the penalized least squares objective function based on the Elastic net penalty is constructed, which can estimate regression coefficients and select important variables, simultaneously. The group effect property of the obtained estimation is proved. The influence of fixed effects is eliminated by the proposed variable selection method and the real model can be identified well even if the covariates are collinear. Secondly, the finite sample properties of the proposed method are accessed by the simulation. The simulation results show that compared with least absolute shrinkage and selection operator (Lasso), adaptive least absolute shrinkage and selection operator (ALasso) and smoothly clipped absolute deviation (SCAD) methods, strongly correlated variables can be better selected by the proposed Elastic net variable selection method. Finally, a real dataset is analyzed by the proposed method.
TibshiraniR. Regression Shrinkage and Selection via the Lasso[J]. J R Stat Soc Ser B Methodol, 1996, 58(1): 267-288. DOI: 10.1111/j.2517-6161.1996.tb02080.x .
[2]
EfronB, HastieT, JohnstoneI, et al. Least Angle Regression[J]. Ann Statist, 2004, 32(2): 407-499. DOI:10.1214/009053604000000067 .
[3]
ZouH. The Adaptive Lasso and Its Oracle Properties[J]. J Am Stat Assoc, 2006, 101(476): 1418-1429. DOI: 10.1198/016214506000000735 .
[4]
FanJ Q, LiR Z. Variable Selection via Nonconcave Penalized Likelihood and Its Oracle Properties[J]. J Am Stat Assoc, 2001, 96(456): 1348-1360. DOI: 10.1198/016214501753382273 .
[5]
ZouH, HastieT. Regularization and Variable Selection via the Elastic Net[J]. J R Stat Soc Ser B Stat Methodol, 2005, 67(2): 301-320. DOI: 10.1111/j.1467-9868.2005.00503.x .
ShuS K, LiL. Multi-factor Stock Selection Strategy Based on Elastic Net Punishment[J]. Stat Decis, 2021, 37(16): 157-161. DOI: 10.13546/j.cnki.tjyjc.2021.16.036 .
WangW S, SongJ H. Open-end Fund Investment Performance Based on Elastic Net Quantile Regression[J]. J Appl Stat Manag, 2020, 39(4): 721-733. DOI: 10.13860/j.cnki.sltj.20191021-003 .
[10]
HuangD X, LiC H, QinC Y, et al. The Application and Property of Elastic Net Procedure for Partially Linear Models[J]. Chin Q J Math, 2020, 35(3): 290-301. DOI: 10.13371/j.cnki.chin.q.j.m.2020.03.004 .
TangL Z, LiY J, ZhaoL J. Study on the Bayesian Elastic Net Quantile Regression for Panel Data: Methods and Applications[J]. Stat Res, 2020, 37(3): 94-113. DOI: 10.19343/j.cnki.11-1302/c.2020.03.008 .
[13]
ShiX Y, LiangB, ZhangQ. Post-selection Inference of Generalized Linear Models Based on the Lasso and the Elastic Net[J]. Commun Stat Theory Meth, 2022, 51(14): 4739-4756. DOI: 10.1080/03610926.2020.1821892 .
ZouJ J, YangY P, ZhaoP X. The Elastic Net Variable Selection for Linear Mixed-effects Models Based on the Orthogonal Projection[J]. Mathematica Applicata, 2024, 37(2): 547-553. DOI: 10.13642/j.cnki.42-1184/o1.2024.02.004 .
LiY, ZengX B. Research on Variable Selection Method of Penalized Likelihood for Panel Data Model[J]. Stat Res, 2014, 31(3): 83-89. DOI: 10.19343/j.cnki.11-1302/c.2014.03.012 .
ZhuL R, HuC Z, LuoY X. A Penalized Composite Quantile Regression Method for Panel Data Model[J]. Stat Decis, 2022, 38(13): 40-45. DOI: 10.13546/j.cnki.tjyjc.2022.13.008 .
[21]
LiG R, LianH, LaiP, et al. Variable Selection for Fixed Effects Varying Coefficient Models[J]. Acta Math Sin Engl Ser, 2015, 31(1): 91-110. DOI: 10.1007/s10114-015-3159-2 .
[22]
ArellanoM. Panel Data Econometrics[M]. Oxford: Oxford University Press, 2003.
[23]
LiN, YangH, YangJ. Nonnegative Estimation and Variable Selection via Adaptive Elastic-net for High-dimensional Data[J]. Commun Stat Simul Comput, 2021, 50(12): 4263-4279. DOI: 10.1080/03610918.2019.1642484 .