ABSTRACT
This chapter provides an introduction to smoothing techniques and how to use those techniques within a Generalized additive models for location scale and shape (GAMLSS) model. It introduces univariate penalized smoothing techniques. The chapter explains the penalized approach to smoothing. It describes the GAMLSS smoothing additive terms including multivariate terms. It discusses univariate and multivariate smoothing techniques and how they are applied within the GAMLSS framework. Both univariate and multivariate smoothers can be used as additive terms within a GAMLSS model formula. The rpart() function of package rpart fits decision trees, and the gamlss.add function tr() provides an interface to it. The earth package in R builds regression models using the techniques in the paper "Multivariate Adaptive Regression Splines" (MARS). Earth is an acronym for "Enhanced Adaptive Regression Through Hinges", used because the term MARS is trademarked.
