High dimensional censored quantile regression (Q1747740)

For the logarithm of a survival time which may be censored, inference of $p$-dimensional unknown coefficients in a linear quantile regression model for its conditional quantile function is considered in this paper. When the number of variables in the model is much larger than the sample size, procedures based on Lasso type penalties and their weighted versions are proposed to simultaneously select the variables which have influences on the target segment of the distribution of the log-survival time and estimate the regression coefficients for the selected variables. Theoretical properties of the proposed procedures are derived, simulation studies confirmed theoretical findings, and an application to real data illustrates the practical utilities of the proposed procedures.