Generalized Additive Models (a.k.a., GAM) comprises a collection of nonparametric and semi-parametric regression techniques for exploring relationships between response and predictor variables. The peculiarity of GAM is that there is no need to make any a priori assumption on the functional form linking the two set of variables, as these relationships are modeled with smooth functions. GAM have recently found wide applications in ecological analysis and particularly in fisheries.
The short course “Applied Generalized Additive Models” aims to introduce the generalized additive regression techniques to ecologists. The main goal of the course is to enable the participants to apply GAM to their data, so they are encouraged to bring their own data and hypothesis to investigate. Topics to be discussed in the short course include:
- Smoothing, Splines
- Generalized additive models (GAM)
- Cross validation (CV)
- Generalized cross validation (GCV)
- Fitting GAM
- Testing for non-additivity
- Model building: partly linear models, transforming the response and the predictors,
- dummy variables, comparing GAMs
- Modeling interactions: continuous interactions, threshold generalized additive models,
- genuine cross validation
The methodologies will be illustrated with ecological data using the mgcv library of R. The short course will be accompanied by a set of lecture notes on GAM and on R. The class is primarily direct to biologists. Some prior knowledge of the participants on regression analysis (linear and nonlinear) is a prerequisite. Although some theoretical background will be covered during the lectures, this class is not meant for statisticians, as it will focus mainly on applications and practical examples. Some familiarity with R (or Splus) programming is advisable; however it is not a requirement as the necessary software codes will be covered during the lectures.