基于纵向数据的超高维特征筛选

Feature Screening for Ultrahigh Dimensional Longitudinal Data

实际问题研究中常常面临复杂数据,其中超高维数据和纵向数据常见于医学、经济学等大数据领域.基于超高维纵向数据的结构特征,推广确定独立筛选SIS(Sure Independence Screening)方法,构造了基于纵向数据组内相关结构的边际特征筛选方法,对超高维问题进行筛选降维,并从理论上证明了所提出降维筛选过程满足确定性筛选性质,从数值模拟上研究了其有限样本性质.

Complex data are often encountered in the study of practical problems,in which ultrahigh dimensional data and longitudinal data are widely faced in the fields of big data such as medicine,economics,and so on. This paper generalizes the sure independence screening（ SIS） method based on the structural property of the ultrahigh dimensional longitudinal data. A marginal feature screening procedure is constructed based on the intra group correlation structure of longitudinal data which can select important predictors and reduce the dimension for ultrahigh dimensional problems.The proposed dimensionality reduction screening process is proved theoretically to satisfy the sure independence screening property and the finite sample property is also examined through numerical studies.