摘要：This paper discusses the selection of the smoothing parameternecessary to implement a penalized regression using a nonconcave penaltyfunction. The proposed method can be derived from a Bayesian viewpoint,and the resultant smoothing parameter is guaranteed to satisfy the sucientconditions for the oracle properties of a one-step estimator. The results ofsimulation and application to some real data sets reveal that our proposalworks eciently, especially for discrete outputs.
