惩罚函数为∑|βj|^γ,γ〉0的桥回归,作为一类特殊的惩罚回归方法,已经被很多文献研究过.本文分别给出了估计系数的压缩大小与γ在γ≥1和0〈γ〈1两种不同情况下取值的一些理论结果,并通过模拟来验证压缩估计的估计效果.
Bridge regression, a special family of penalized regressions of a penalty function ∑|βj|^γ with γ〉0, has been studied in many literatures. In this paper, we provide some theoretical results of how the shrinkage rule changing with γ under two settings: γ≥1 and 0〈γ〈1, respectively. Simulation results are conducted to evaluate the performance of the proposed method.