2017年06月22日，星期四下午15:30在大学城广州大学行政东楼前座412会议室。香港理工大学赵兴球教授为各位学子带来了精彩的学术讲座--《Subgroup Analysis in Censored Linear Regression》。
Treatment heterogeneity problems with censored survival data are considered. To solve the problems, we propose a censored and heterogeneous linear regression model, in which the treatment effects are subject-specific and belong to different subgroups without knowing the group membership of the individuals in advance and develop a concave fusion penalized method for subgroup analysis. The proposed inference procedure can automatically specify the subgroup structure and estimate the subgroup-dependent treatment effects simultaneously. In particular, the resulting estimators enjoy the oracle property. To implement the proposed procedure, we combine the Buckley-James iterative procedure with the alternating direction method of multipliers algorithm and demonstrate its convergence. Simulation studies indicate that the new method works well and a real data example is provided for illustration.