Generalised Additive Models

Dozierende Prof. Simon Wood, University of Bath, UK
Abschluss Teilnahmebestätigung

Advanced R users from all professional groups. (For introductory R courses please revisit the course list.)

  • CHF 600.- für Angehörige der UZH/ETH und assoziierter Institute
  • CHF 800.- für Alumni der UZH/ETH, Angehörige anderer Universtitäten, Forschungseinrichtungen und Ämter des Bundes oder der Kantone, non-profit Organisationen
  • CHF 1200.- für Firmen
Kurssprache Englisch

Generalized additive models (GAMs) are generalized linear regression models in which the linear predictor depends on smooth functions of predictor variables. Such models offer very flexible model specification, making them well suited to situations in which there are complicated spatial or temporal dependencies in the data, that are not easily dealt with parametrically.

This course will concentrate on the approach to GAMs exemplified in R package mgcv, in which the smooth terms are represented using reduced rank smoothing splines, with associated quadratic penalties, and smoothing parameter selection is achieved via generalized cross validation, REML or similar criteria. The framework developed covers smooth modelling beyond single parameter exponential families, also allowing multivariate additive modelling, Cox proportional hazards models, location scale models, models for ordered categorical data, as well as Tweedie and negative binomial models.

The course will be practically and computationally focused, starting with the basics of basis-penalty smoothing, before moving on to look at the representation and estimation of GAMs, and the R package mgcv. The wide range of smooth functions that can form model components will then be covered, followed by material on model checking and selection. The course will finish by looking at some more advanced topics, such as posterior simulation, functional data analysis and models well outside the simple exponential family class. A substantial proportion of the time will be devoted to computer exercises.

For all Zurich R Courses participants should bring their own laptops to the course and will be informed by email in advance which packages they need to install.

Daten January 28-29, 2016
Anmeldeschluss: 20.12.2015
  Nach der Anmeldung erhalten Sie zunächst eine kurze automatische Anmeldebestätigung per Email. Wenn Sie diese Email erhalten haben, sind Sie erfolgreich und verbindlich zum Kurs angemeldet. Die schriftliche Rechnung wird aus administrativen Gründen erst ca. zwei Wochen vor Kursbeginn verschickt.