Model validation and error estimation in multi-block partial least squares regression

While validation of Partial Least Squares Regression (PLSR) models has been discussed extensively, validation tools that are tailored to Multi-block Partial Least Squares Regression (MBPLSR) have not been discussed in literature yet. This paper introduces validation tools for estimating predictive ability and model stability in MBPLSR models on block level and on global level. Predictive ability on the block level and global level are estimated by calculating the predictive power of block and global parameters. Model stability is estimated by checking the stability of block model parameters and global parameters. By comparing error plots for model stability and predictive ability the user can decide on the number of component to be used. The number of components to be chosen depends on the data set and the purpose of the investigation. ⺠Multi-block PLSR detects and displays y-relevant co-variation patterns in the data. ⺠Tools to assess the relative importance of the individual blocks are proposed. ⺠Prediction ability of each block is visualized. ⺠Model stability in individual blocks is estimated. ⺠Optimal rank in global and block models is determined.