Non-linear Multi-block Partial Least Squares via Univariate and Bivariate B-splines

In the last decade much effort has been spent modelling dependence among variables to probe the relationships between response variables and predictor ones observed at different occasions/spaces/times. In this paper we propose a nonlinear generalization of multi-block Partial Least Squares using multivariate additive splines. We show the method performance on real sensory data sets.