Study On The Applications Of Variable Selection In Ultrahigh Dimensional Regression

In this thesis,we give the RCV-Elastic algorithm by using the variance estimation method of Refitted cross-validation in the Elastic Net method.This algorithm have stability as the Elastic Net method,and performed well in variance estimation as a result of Refitted cross-validation.Numerical experiments show that the variance obtained from the RCV-Elastic algorithm perform more oracle than the naive two stage estimators.The real case analysis show the progress of variance selection in ultrahigh dimension regression.