PLS regression based on sure independence screening for multivariate calibration

Abstract

By employing the simple but effective principle sure independence screening, a novel strategy for selecting the variables, named PLS regression combined with sure independence screening (PLSSIS), is developed. The PLSSIS algorithm combines the sure screening and latent variables method, which can fastly and efficiently deal with the high dimensional collinear data. Under the sure independence screening, the reduced submodel contains all the variables in the true model with probability tending to one. Our study shows that better prediction is obtained by PLSSIS compared with PLS modeling and moving window partial least squares (MWPLS).