带有时间序列的变系数EV模型的变量选择

Variable selection for varying coefficient EV models with time series

在模型误差是时间序列时,利用B样条逼近和SCAD惩罚函数对变系数EV模型进行变量选择。选择合适的调整参数,偏差修正的变量选择能够同时选择有效的变量和估计非零的光滑系数函数。最后证明了变量选择的相合性,同时它也满足变量选择的Oracle性质——稀疏性。

This paper focuses on variable selection for varying coefficient EV models when model error is a time series .We use B-spline basis and the SCAD penalty function to select .With appropriate tuning parameters , the proposed method can select significant variables and estimate the nonzero smooth coefficient functions simultaneously .We establish the consistency of the variable selection procedure , further the optimal convergence rate of the regularized estimator can be derived .