基于距离相关的超高维竞争风险数据特征筛选

Feature screening for ultra-high dimensional competing risks data based on distance correlation

针对超高维竞争风险数据,提出并研究了一种新的基于距离相关的特征筛选方法.从理论和数值模拟两方面对新提出的方法进行了研究.理论上证明了新方法的确定性筛选性质.数值上的结果表明,新提出的方法比现有筛选方法有更好的表现,并将新方法应用到了一个实际问题.

This paper proposes a feature screening method based on distance correlation for the ultra-high dimensional competing risks data. Theoretically, sure screening property is well established. In addition, some numerical studies are conducted to assess the finite-sample property. The numerical results show that the newly proposed method has better performances than the existing screening methods. At last, this new procedure is applied to a real dataset.