Robust Estimation and Variable Selection for Partially Linear Panel Data Models with Fixed Effects

Abstract

The paper proposes a new composite quantile regression estimation and variable selection procedures for partially linear panel data models with fixed effects. By combining forward orthogonal derivations transform with B-spline approximations, we first develop a semiparametric composite quantile regression procedure. The main advantage of the proposed method lies in its robustness compared to the least-squares-based method, especially for many non-normal errors. Under some regularity conditions, we establish the asymptotic properties of the resulting estimators. To achieve sparsity with high-dimensional covariates, we further propose adaptive Lasso penalized composite quantile regression estimation for variable selection in partially linear panel data models, and establish the oracle property under some regularity conditions. Simulation studies and a real data analysis are provided to assess the finite-sample performance of the proposed procedures.

Keywords

B-spline / Composite quantile regression / Panel data / Partially linear model / Variable selection / 62G08 / 62G20

References

[1] ArellanoMPanel Data Econometrics, 2003, Oxford. Oxford University Press.

[2] Arteaga-MolinaLA, Rodriguez-PooJM. Empirical likelihood based inference for a categorical varying-coefficient panel data model with fixed effects. J. Multivar. Anal., 2019, 173: 110-124

[3] BradicJ, FanJ, WangW. Penalized composite quasi-likelihood for ultrahigh dimensional variable selection. J. R. Stat. Soc. Series B (Stat. Methodol.), 2011, 73(3): 325-349

[4] ChenJ, GaoJ, LiD. Estimation in partially linear Single-Index panel data models with fixed effects. J. Bus. Econ. Stat., 2013, 31(3): 315-330

[5] ChenR, LiG, FengS. Testing for covariance matrices in time-varying coefficient panel data models with fixed effects. J. Korean Stat. Soc., 2020, 49(1): 82-116

[6] FanJ, LiR. Variable selection via nonconcave penalized likelihood and its oracle properties. J. Am. Stat. Assoc., 2001, 96(456): 1348-1360

[7] FengS, LiG, PengH, TongT. Varying coefficient panel data model with interactive fixed effects. Stat. Sin., 2021, 31(2): 935-957

[8] GuoJ, TangM, TianM, ZhuK. Variable selection in high-dimensional partially linear additive models for composite quantile regression. Comput. Stat. Data Anal., 2013, 65: 56-67

[9] Hamiye BeyaztasB, BandyopadhyayS. Robust estimation for linear panel data models. Stat. Med., 2020, 39(29): 4421-4438

[10] HeB, HongX, FanG. Block empirical likelihood for partially linear panel data models with fixed effects. Stat. Prob. Letter, 2017, 123: 128-138

[11] HeX, ShiP. Convergence rate of B-spline estimators of nonparametric conditional quantile functions. J. Nonparam. Stat., 1994, 3: 299-308

[12] HuangJ, HorowitzJL, WeiF. Variable selection in nonparametric additive models. Ann. Stat., 2010, 38(4): 2282-2313

[13] HuangX, LinZ. Local composite quantile regression smoothing: a flexible data structure and cross-validation. Economet. Theor., 2021, 37(3): 613-631

[14] JiangR, QianW, ZhouZ. Single-Index composite quantile regression with heteroscedasticity and general error distributions. Stat. Pap., 2016, 57(1): 185-203

[15] JiangR, YuK. Single-index composite quantile regression for massive data. J. Multivar. Anal., 2020, 180104669

[16] JiangXJ, JiangJ, SongX. Oracle model selection for nonlinear models based on weighted composite quantile regression accelerated failure time model. Stat. Sin., 2012, 22: 1479-1506

[17] KnightK. Limiting distributions for $L_1$ regression estimators under general conditions. Ann. Stat., 1998, 26(2): 755-770

[18] KimM. Quantile regression with varying coefficients. Ann. Stat., 2007, 35(1): 92-108

[19] LaiP, LiG, LianH. Semiparametric estimation of fixed effects panel data single-index model. Stat. Probab. Letter., 2013, 83(6): 1595-1602

[20] LamarcheC. Robust penalized quantile regression estimation for panel data. J. Econ., 2010, 157(2): 396-408

[21] LiG, LianH, LaiP, PengH. Variable selection for fixed effects varying coefficient models. Acta Math. Sinica, 2015, 31(1): 91-110

[22] LiS, WangK, RenY. Robust estimation and empirical likelihood inference with exponential squared loss for panel data models. Econ. Lett., 2018, 164: 19-23

[23] LiuH, YangH, XiaX. Robust estimation and variable selection in censored partially linear additive models. J. Korean Stat. Soc., 2017, 46(1): 88-103

[24] MunnellA. Why has productivity growth declined? Productivity and public investment. New Engl. Econ. Rev., 1990, 1990: 3-22

[25] SchumakerLLSpline functions, 1981, New York. Wiley.

[26] SuL, UllahA. Profile likelihood estimation of partially linear panel data models with fixed effects. Econ. Lett., 2006, 92: 75-81

[27] StoneCJ. Additive regression and other nonparametric models. Ann. Stat., 1985, 13(2): 689-705

[28] StoneCJ. Optimal global rates of convergence for nonparametric regression. Ann. Stat., 1982, 10(4): 1040-1053

[29] TangL, ZhouZ, WuC. Weighted composite quantile estimation and variable selection method for censored regression model. Stat. Prob. Letter., 2012, 82(3): 653-663

[30] TianYZ, ZhuQ, TianM. Estimation of linear composite quantileregression using EM algorithm. Stat. Prob. Letter., 2016, 117: 183-191

[31] WangH, LiR, TsaiCL. Tuning parameter selectors for the smoothly clipped absolute deviation method. Biometrika, 2007, 94(3): 553-568

[32] WangKN, LiS, ZhangB. Robust communication-efficient distributed composite quantile regression and variable selection for massive data. Comput. Stat. Data Anal., 2021, 161107262

[33] WeiF. Group selection in high-dimensional partially linear additive models. Brazil. J. Prob. Stat., 2012, 26(3): 219-243

[34] YangH, LiuH. Penalized weighted composite quantile estimators with missing covariates. Stat. Pap., 2016, 57(1): 69-88

[35] ZouH, YuanM. Composite quantile regression and the oracle model selection theory. Ann. Stat., 2008, 36(3): 1108-1126

[36] ZhaoWH, LianH, SongX. Composite quantile regression for correlated data. Comput. Stat. Data Anal., 2017, 109: 15-33

[37] ZhangJ, FengS, LiG, LianH. Empirical likelihood inference for partially linear panel data models with fixed effects. Econ. Lett., 2011, 113(2): 165-167

[38] ZhangR, LvY, ZhaoW, LiuJ. Composite quantile regression and variable selection in single-index coefficient model. J. Stat. Plann. Inference, 2016, 176: 1-21

Funding

Natural Science Foundation of Chongqing(cstc2021jcyj-msxmX0079)

Humanities and Social Sciences Program of Chongqing Education Commission(21SIGH118)

National Social Science Foundation of China(18BTJ035)

National Natural Science Foundation of China(12271046, 12131006, 11971001)

RIGHTS & PERMISSIONS

School of Mathematical Sciences, University of Science and Technology of China and Springer-Verlag GmbH Germany, part of Springer Nature