SCGLR-An R Package for Supervised Component Generalized Linear Regression

The objective of this paper is to present an R package, SCGLR, implementing a new PLS regression approach in the multivariate generalized linear framework. The method allows the joint modeling of random variables from different exponential family distributions, searching for common PLS-type components. We discuss several of the functions in the package focusing in particular on the two main ones: scglr and scglrCrossVal. The former constructs the components and performs the parameter estimation, while the latter selects the approriate number of components by cross-validation. The package is illustrated on an appropriate ecological dataset through which we aim at predicting the abundance of multiple tree genera given a large number of geo-referenced environmental variables.