Netherlands, Nijmegen

Practical Mixed Effect Regression Modelling for Psychology and Language Science

when 5 August 2019 - 9 August 2019
language English
duration 1 week
credits 2 EC
fee EUR 450

The course covers mixed effect modelling; a cutting-edge regression method for cognitive and social scientists. In lectures and hands-on tutorials using real data sets, students will run, fit, interpret and visualize models using R.

This course is a practical guide to mixed effect modelling, a cutting-edge regression method important for cognitive and social science. The course is structured as 50% lecture and 50% hands-on practice where students are guided through tutorials in R, a popular open-source programming language and statistical interface. Students will design code and analyses to examine their choice of data (provided by us, or brought to class by the student), with the aim of providing useful, concrete skills that will translate from the classroom to academic or professional expertise.

The course begins with an overview of the principles of regression and basic programming techniques in R: Students will learn how mixed-effect modelling relates to other regression models, its underlying assumptions, and how to implement simple analyses using a common and useful programming language.



Next, we will cover differences between fixed and random effect models and show how to diagnose what models fit the data the best, pairing this with hands-on tutorials. We discuss why the mixed effect modelling technique is superior to traditional repeated measures ANOVA. In this process, students will gain a deep understanding of what various regression methods do, as well as how to implement and use mixed-effect modelling for their own work.

We will also cover practicalities about how to analyse different sorts of predictors (contrast coding and transformations) and dependent measures (linear vs generalized mixed effect models). Understanding these properties of models is crucial for understanding how to interpret a model’s results, and gaining these skills will provide students with the ability to create analyses to efficiently answer any question of interest.

Finally, we will cover experimental power—what it is, new ways of thinking about it, and how to run power analyses using mixed effect models. Calculating statistical power allows scientists to design studies that have the right number of observations to reliably test questions of interest; as such, issues of experimental power are entangled with understanding the replicability and reliability of the scientific literature. In the past, power analyses have often been cumbersome and full of assumptions, but with the advent of mixed effect models, designing powerful studies has never been easier. We will demonstrate how to do this and how to avoid misleading or underpowered null results in the future.

Course leader

Laurel Brehm
Research Staff
Max Planck Institute for Psycholinguistics

Phillip Alday
Research Staff
Psychology of Language
Max Planck Institute for Psycholinguistics

Target group

Advanced Bachelor
Master
PhD

The course is designed for students in psychology and language science who use repeated-measures designs in empirical research. The target audience is research students beginning on their first research projects, with at least passive knowledge of probability and statistics.

Course aim

After this course you are able to:

Use & implement mixed effect models in R,
Report mixed effect models for publication,
Understand statistical power for effective experimental designs.

Fee info

EUR 450: Normal fee

Scholarships

€ 405 early bird discount – deadline 1 March 2019 (10%)
€ 383 partner + RU discount (15%)
€ 338 early bird + partner + RU discount (25%)

Register for this course
on course website