Variable Selection Based On Generalized Measures Of Correlations For High Dimensional Model

Variable selection plays an important role in dealing with the high dimensional data Sure independence screening is a variable selection method that was proposed by Fan and Lv (2008). We note that the idea of SIS is based on Pearson’s correlation to select significant variables. However, Pearson’s correlation coefficient has a lot of lim-itations. For example, it can only measures the linear and nonsymmetric correlation. Zheng et al.(2012) proposed an index which can measure nonlinear or asymme-try correlation that is the generalized measures of correlation(GMC). Motivated by the idea of SIS, in this paper we propose a method based on generalized measure of correlation(GMC)rather than the Pearson’s correlation to deal with high dimensional data, called GMC-SIS. Simulations and a real data analysis are carried out. Smulation results will show the advantages and disadvantages.