Introduction

This introduction presents an overview of the key concepts discussed in the subsequent chapters of this book. The book describes some basic scatterplot smoothers, like the running mean, locally-weighted running-line, kernel and cubic-spline smoothers, and also looks briefly at smoothers for multiple predictors. It discusses some important issues such as how to choose the smoothing parameters for a given smoother, and how to make inferences about the fitted smooth. The book also discusses somewhat more technical in nature, and places additive models and backfitting on more solid theoretical ground. It provides applications of the additive model to a few nonstandard situations, for example, models for survival, case-control and ordered categorical response data. The book explores some special topics including the modelling of interactions, resistant fitting and model selection. It explains the extension of additive models to binary data, for which the response takes on only two values such as survived or died.