Tenth Summer School on Statistical Methods for Linguistics and Psychology

Advanced methods in frequentist statistics with Julia

Authors

Phillip Alday

Reinhold Kliegl

Published

2026-06-21

This site provides materials for the Advanced frequentist methods stream of the Summer School on Statistical Methods to be held at the University of Potsdam, 24–28 August 2026. The stream focuses on fitting, evaluating, and interpreting linear and generalized linear mixed-effects models using the Julia programming language and the MixedModels.jl package.

1 How to use these materials

Every page is a complete, runnable analysis. You can engage with them two ways:

  • Read along. The rendered website shows all code and its output, so you can follow the reasoning without running anything. Good for a first pass.
  • Run locally. Clone the repository and execute the code yourself — change things, break things, and do the exercises at the bottom of each core page. This is where the learning happens. Start with Setup: installing the tools.

Most pages in the core path open with a Learning objectives box (what you’ll be able to do afterwards) and a Before you start box (which pages to read first and which datasets are used), so you can navigate and pace yourself without an instructor in the room.

TipI just need to look something up

The Reference & Appendix section is lookup material, not a lesson sequence: a tour of useful Julia packages, how to save and load data with Arrow, and package management for reproducible environments. Dip in as needed.

2 Suggested path

  1. Start Here — install the tools, meet the instructors, and skim the dataset catalog.
  2. Foundations: Linear Mixed Models — fit and interpret your first mixed model (sleepstudy) and learn to recognize singular fits.
  3. Seeing Your Model — shrinkage plots and multi-panel graphics; diagnostic tools you’ll reuse everywhere.
  4. Building Models — random-effects specification and selection for designed and large-scale experiments.
  5. Contrast Coding & Transformations — make your coefficients answer the questions you actually care about (transformations, contrast schemes).
  6. Inference & Uncertainty — bootstrap, profiling, and power simulation.
  7. Generalized Linear Mixed Models — extending to non-Gaussian responses.
  8. Capstone Worked Examples — full analyses that pull the pieces together: the Emotikon project and the visual-attention experiments.

3 Dependency map

The core path is roughly linear, but not strictly: once you have the Foundations, several sections can be read in any order. The arrows show what each section assumes you’ve already seen.

flowchart TD
    F["Foundations:<br/>Linear Mixed Models"]
    V["Seeing Your Model:<br/>Visualization & Diagnostics"]
    B["Building Models:<br/>Specification & Random Effects"]
    C["Contrast Coding<br/>& Transformations"]
    I["Inference &<br/>Uncertainty"]
    G["Generalized Linear<br/>Mixed Models"]
    X["Capstone<br/>Worked Examples"]

    F --> V
    F --> B
    F --> C
    F --> G
    B --> I
    V --> X
    C --> X
    I --> X

4 Datasets at a glance

A handful of datasets recur across the course. The dataset catalog describes each one — its design, variables, citation, and how to load it. Here is where each appears:

Dataset Used in
sleepstudy sleepstudy, singularity, transformations, speed metrics, bootstrap, GLMM
kb07 shrinkage plots, bootstrap, profiling, power simulation
fggk21 (Emotikon) contrast coding, AoG plots, Emotikon capstone, transform check
kwdyz11 contrast coding, visual-attention capstone
kkl15 replication & model reduction
mrk17 model specification

5 Getting started

New here? Head to Setup: installing the tools to get git, Julia, and Quarto working before the course, then read About for the course framing and instructor bios.

This page was rendered from git revision e563d55 using Quarto 1.9.38 and Julia 1.12.6.

Back to top