2017-06-16 | 在超维变系数模型下的有效的变量选择
Feature screening in ultrahigh-dimensional varying coefficientmodels is a crucial statistical problem in economics,genomics and etc. Existing methods suffer in the casesof multiple index variables and group predictor variables.Moreover, current methods can not handle nonlinear varyingcoefficient models which is possible in reality. To deal withthose scenarios efficiently in real life, we develop a screeningprocedure for ultrahigh-dimensional varying coefficientmodels utilizing conditional distance covariance (CDC). Extensivesimulation studies and two real economic data exampleshave shown the effectiveness and the flexibility ofour proposed methods.
Dr Chen Xin is an assistant professor from Dept. of Statistics and Applied Prob. in National University of Singapore. He got his bachelor degree from Nankai University and his Ph.D degree from University of Minnesota. His research interests include dimension reduction, variable selection and high dimensional analysis.