arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1302.4244math.STstat.TH
收藏

高维分位数回归中的突变问题

Quantile regression in high-dimension with breaking

Gabriela Ciuperca

摘要:

本文考虑了高维线性回归模型,其中预测变量在未知时刻(称为变点)对响应变量的影响可以改变。此外,考虑了重尾误差的特殊情况。在这种情况下,不能使用带有LASSO或自适应LASSO惩罚的最小二乘法,因为理论假设不成立或估计量不稳健。然后,分位数模型与SCAD惩罚或中位数回归与LASSO型惩罚允许,在同一时间,估计每个段上的参数,并消除不相关的变量。我们证明了,对于这两种惩罚估计方法,预言性质不受变点估计的影响。给出了这两种方法对变点和回归参数的估计的收敛速度,Monte-Carlo模拟说明了这两种方法的性能。

英文摘要:

The paper considers a linear regression model in high-dimension for which the predictive variables can change the influence on the response variable at unknown times (called change-points). Moreover, the particular case of the heavy-tailed errors is considered. In this case, least square method with LASSO or adaptive LASSO penalty can not be used since the theoretical assumptions do not occur or the estimators are not robust. Then, the quantile model with SCAD penalty or median regression with LASSO-type penalty allows, in the same time, to estimate the parameters on every segment and eliminate the irrelevant variables. We show that, for the two penalized estimation methods, the oracle properties is not affected by the change-point estimation. Convergence rates of the estimators for the change-points and for the regression parameters, by the two methods are found. Monte-Carlo simulations illustrate the performance of the methods.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑