Sparse Dimension Reduction for Survival Data
In this paper, we consider the estimation and variable selection of the sufficient dimension reduction space for survival data through a new combination of L1 penalty with the outer product of gradient method (OPG, Xia et al., 2002), called SHOPG hereafter. The SH-OPG can exhaustively estimate the central subspace and select the informative covariates simultaneously. On the other hand, the estimated directions remain orthogonal automatically after ticking noninformative regressors out. The efficiency of SH-OPG is verified through extensive simulation studies and real data analysis.
censored data hazard function variable selection dimension reduction
Changrong Yan Dixin Zhang
Center of Research of Finance Econometrics and Risk Management Department of Finance and Insurance,Economics School Nanjing University,China
国际会议
Second Joint Biostatistics Symposium(第二届生物统计国际研讨会2012)
北京
英文
461-479
2012-07-08（万方平台首次上网日期，不代表论文的发表时间）