摘要
文章提出一种同时对固定效应和随机效应施加双SCAD惩罚的随机效应分位回归模型。使用两步迭代算法求解参数,并用SIC和GACV准则对惩罚参数进行选取,在这两种准则下进行了蒙特卡洛模拟比较,分别模拟了在误差不同和变量相关系数不同的情况下的估计和选择情况。基于GACV准则将提出的方法在不同情形下与其他两种方法进行模拟对比,最后用实例数据进行验证。结果表明:该模型系数估计和变量选择的结果良好,且提出的方法非常适用于稀疏模型。
This paper proposes a quantile regression model of random effects with double SCAD punishment for both fixed effects and random effects. This paper uses the two-step iterative algorithm to solve the parameters, and selects the penalty parameters by SIC and GACV criteria, with Monte Carlo simulation carried out under these two criteria, and the estimation and selection respectively simulated under different errors and variable correlation coefficients. And then, based on the GACV criterion, the paper compares the proposed method with the other two methods in different situations. Finally, example data is used to make verification. The results show that the coefficient estimation and variable selection results of the model are good, and the proposed method is very suitable for sparse models.
作者
任雪妮
罗幼喜
Ren Xueni;Luo Youxi(School of Science,Hubei University of Technology,Wuhan 43006&China)
出处
《统计与决策》
CSSCI
北大核心
2021年第18期9-13,共5页
Statistics & Decision
基金
国家社会科学基金资助项目(17BJY210)。
关键词
双SCAD惩罚
随机效应
分位回归
迭代算法
double SCAD punishment
random effects
quantile regression
iterative algorithm
