Adaptive group bridge selection in the semiparametric accelerated failure time model (Q2293393)

The accelerated failure time model is considered with right censored data. In the literature, the group penalized method is studied, which is able to remove unimportant groups of covariates but cannot effectively remove unimportant covariates within the important group. The paper under review proposes the adaptive group bridge penalized Stute's weighted LS estimators and shows their powerful oracle properties. The dimension of covariates is allowed to be larger than the sample size, which is applicable for the high-dimensional genomic data. Numerical studies indicate that the proposed approach can effectively conduct group and within-group individual variable selection simultaneously, even with high censoring rates in high-dimensional data, and outperforms the existing group selection methods.