Welcome to the mgcv course webpage

A course! On mgcv and generalized additive models.

This site has slides and examples from the course as well as other useful information.

The course is based on a course given by Dave Miller, Eric Pedersen and Gavin Simpson at the Ecological Society of America Annual Meeting 2016. That course’s website can be found here.

Course overview

To address the increase in both quantity and complexity of available data, ecologists require flexible, robust statistical models, as well as software to perform such analyses. This workshop will focus on how a single tool, the mgcvpackage for the R language, can be used to fit models to data from a wide range of sources.

mgcv is one of the most popular packages for modelling non-linear relationships. However, many users do not know how versatile and powerful a tool it can be. This workshop will focus on teaching participants how to use mgcv in a wide variety of situations (including spatio-temporal, zero-inflated, heavy-tailed, time series, and survival data) and advanced use of mgcv (fitting smooth interactions, seasonal effects, spatial effects, Markov random fields and varying-coefficient models).

The workshop will give paricipants an understanding of:

  • practical elements of smoothing theory, with a focus on why they would choose to use different types of smoothers
  • model checking and selection
  • the range of modelling possibilities using mgcv.

Participants will be assumed to be familiar with the basics of R (loading/manipulating data, functions, and plotting) and regression in R (lm() and glm()). The organizers have extensive practical experience with ecological statistics and modelling using mgcv.