线性EIV模型参数的LASSO估计方法

The LASSO estimation method for linear EIV model parameters

基于结构风险最小化原则,提出了可以实现高维数据降维的线性EIV模型参数的LASSO估计(LE)方法,并给出了其数值解的迭代算法。为说明LE方法的有效性,通过实证与WTLS、LS两种方法进行了对比分析。结果表明,LE方法能够明显提高预测精度,具有更强的泛化能力,同时可以实现变量选择,达到高维数据降维的目的。

In this paper,considering the principle of structural risk minimization,the LASSO estimation method for linear EIV model parameters(LE),which can achieve dimensionality reduction,is proposed,and its numerical solution with iterative algorithm is given.In order to illustrate the effectiveness of the LE method,it was campared with WTLS and LS methods.The results showed that the LE method proposed in this paper can significantly improve the prediction accuracy,and achieve high generalization and variable selection to fulfill the purpose of dimensionality reduction of high-dimensional data.