截面相关面板模型的去偏Lasso估计

研究目标:克服截面相关面板数据模型参数估计基准CCE方法随着自变量增加而参数估计置信水平下降的缺点。研究方法:在使用因变量与自变量的截面平均去除因子结构之后使用去偏Lasso的方法代替最小二乘方法得到参数估计值以及置信区间。研究发现:通过模拟发现去偏Lasso-CCE能够很好地解决自变量过多导致基准CCE参数估计方法置信度降低的问题,有效弥补了基准CCE方法的不足。研究创新:发现变量增加导致基准CCE参数估计方法置信度降低、重要解释变量选择出现偏误,提出去偏Lasso-CCE方法解决这一问题并且证明估计量满足一致性和渐近正态性。研究价值:把高维相关理论应用到面板计量模型,拓展了CCE方法的适用性,有利于截面相关面板数据模型的应用研究。

Debiased Lasso for Panel Data Model with Cross-section Dependence

Research Objectives:Overcoming the problem of reduced confidence level with increasing independent variables in cross-dependence panel data model by benchmark CCE method.Research Methods:After using the cross-sectional average of dependent and independent variables removal factor structure,the method of debiased Lasso is proposed to obtain the parameter estimates and confidence intervals.Research Findings:The debiased Lasso CCE can solve the problem that the benchmark CCE method has low confidence due to too many independent variables,which effectively compensates for the shortage of benchmark CCE method.Research Innovations:Increased variables lead to decreasing of confidence level by benchmark CCE method.A debiased Lasso CCE is proposed to solve this problem and the consistency and asymptotic normality of the estimator are proved.Research Value:The application of high-dimensional theory to panel model expands the applicability of CCE method.Beneficial to applied research for cross dependence panel data model.