A Reasearch On FDR Of Correlations Of The Linear Model

Our society are now on a “Big Data” era, that a largevariety of data are being generated from fields, such asgenetics and Internet. How to explore as much information aswe can has drawn attention from more and more people. Lasso isa good method for people to get the correlations betweenvariances in a linear model. It sets some coefficients ofcorrelations to zero, which makes it more powerful thantraditional methods. FDR is a very useful method for us to knowhow accurate one technique can be. However, by Lasso, we cannot know if our decision that the coefficient is zero is trueor not, that is, we can not control its FDR. In this paper, weproposed a procedure of deciding whether the coefficients inlinear model is zero, and we prove both theoretically andnumerically that our procedure can control its FDR, and, in thatsense, a better method than Lasso.