Hello, I am trying to use sparse PLS to analyze the prediction relationship, and I found sPLS can be performed with more variables than samples while analyzing the n>>p problem. However, if p > n, the lasso select at most n variables. Thus, I don’t know why we can select more variables using mixOmics. Thank you in advance.
We are using a soft-thresholding approach to solve the lasso here.
Thanks a lot for your quick reply. I wonder if it is ok that more variables (greater than n) are selected from a mathematics view? Can you recommend some literature for me?