Bootstrap selection of Multivariate Additive PLS Spline models
Abstract
En
Multivariate Additive PLS Splines, in short MAPLSS, are Partial Least-Squares models that study the dependence of a set of responses on spline transformations of the predictor variables which permit to capture additively non linear main effects and interactions. The aim of this paper is to present a way of selecting MAPLSS models through an adaptive incremental selection of training samples by a bootstrap procedure. This approach is attractive in the case of expensive data thus implying to construct efficient models based on small training data sets.
Multivariate Additive PLS Splines, in short MAPLSS, are Partial Least-Squares models that study the dependence of a set of responses on spline transformations of the predictor variables which permit to capture additively non linear main effects and interactions. The aim of this paper is to present a way of selecting MAPLSS models through an adaptive incremental selection of training samples by a bootstrap procedure. This approach is attractive in the case of expensive data thus implying to construct efficient models based on small training data sets.
DOI Code:
�
Keywords:
Bootstrap; PLS regression; B-splines; Design of experiments
Full Text: PDF
