高维广义线性模型的拟似然自适应Lasso估计

Quasi-likelihood adaptive Lasso estimators for high-dimensional generalized linear models

利用惩罚拟似然方法,讨论高维广义线性模型的拟似然自适应Lasso估计。该方法能同时进行变量选择和参数估计。在适当的条件下,证明了所得估计的相合性和Oracle性质,并利用数据模拟和实例分析说明了所提方法的优良性质。

Using the penalized quasi-likelihood method, the adaptive Lasso quasi-likelihood estimators in high-dimensional generalized linear model are discussed. The proposed method can perform variable selection and estimation simultaneously. Under regularity conditions, the consistency and Oracle property of the adaptive Lasso estimator are obtained. These results are examined by several simulation studies and a real data example.