“…The key is to extract sufficient quantity of variables, which can be considered as high-dimensional variables combinations. 26,41 Consider n observations for each cell type t, t = 1, ..., T, and p variables data set (y,X t ), where X t = [x 1t ,x 2t , ..., x pt ], x jt = [x j1t ,x j2t , ...,x jnt ] T is an n × 1 vector for the jth variable, j = 1, ...,p. Ridge and Lasso conduct variable selections based on the following generalized linear model for multinomial response,…”