Jacolien Van Rij, University of Tübingen
Juhani Järvikivi, University of Alberta

This course provides an introduction in how to analyze eye-tracking data using Generalized Additive Mixed Modeling (GAMM; Lin & Zhang, 1999) as implemented in the R package mgcv (Wood, S.N., 2006; 2011). GAMM is nonlinear regression method, which allows analyzing the time course of the gaze data or pupillary response. Although the course focuses on gaze data and pupil dilation data from Visual World Paradigm experiments, GAMMs can also be used for other eye tracking data, such as from reading experiments.

The course consists of three subparts:

  1. The first part focuses on how to prepare the eye-tracking data for analysis, which involves exporting and preprocessing the data. Topics such as blink detection, missing data, time alignment, down sampling, and baseline corrections will be discussed.
  2. The second part introduces GAMMs for the analysis of gaze data. The visualization and interpretation of nonlinear interactions and nonlinear random effects are being discussed.
  3. The third part uses GAMM to analyze pupillary response data and explains how to evaluate GAMM models. Model comparison procedures, summary statistics, and autocorrelation in the residuals are being discussed.

The course combines short lectures with practical hands-on sessions. We will use R for the analysis.

Prerequisites for this course:

Although this course is follow-up on the eyetracking course “Introduction and Data Collection,” knowledge about eye tracking is not required for following this course. So it’s possible to follow this course without taking the introduction eye-tracking course.

  • Basic experience with R: working with data frames, vectors, and lists; using for-loops and aggregate() and/or tapply(). If you do not have experience with R, please consider taking the Introduction to R course.
  • Basic knowledge about linear regression in R: what are regression lines, intercept and slope; how to interpret a model summary; factors versus numeric predictors. Basic knowledge about random effects is not required, but very useful. If you do not have experience with linear regression or linear mixed effects models, please consider taking the Introduction to R course before attending this course.
  • Laptop / computer with R or R studio installed.